Materials Research — the science of battery, magnet & extraction chemistry

HELD viable on cells shipping in two-wheelers, start-stop and ESS. Passenger-EV sodium-ion is a WATCH ITEM only: CATL/Changan announced the A06 sodium variant (Naxtra layered oxide, 175 Wh/kg claim) for mid-2026 (CATL newsroom 2026-02-05; electrive 2026-02-09); as of 2026-09-05 no independent confirmation of customer deliveries and no third-party energy-density test. Both lenses converge; nothing above the cell-level viable claim is admitted.

★ For the kids: Salt-based batteries are cheaper and love the cold; the first family car with one is promised this year, but we have not seen it delivered yet.

Technical: Layered O3/P2 oxides remain air/moisture sensitive; PBA and polyanion branches carry their own records. CATL cites four solved mass-production bottlenecks (moisture, hard-carbon gassing, Al-foil adhesion, self-forming anode) - manufacturer claims.

Frozen prediction: By 2027-03-31 >=2 independent outlets or a Changan/CATL filing confirm customer deliveries of the sodium-ion A06, OR CATL reports cumulative Naxtra shipments >=1 GWh (p=0.60); by 2027-06-30 CATL/Changan or CPCA report >=10,000 cumulative sodium-ion passenger-vehicle deliveries (p=0.5).

HELD at viable (IEA Global EV Outlook 2025: LFP roughly half of global EV battery output in 2024). The 'LFP relieves the cobalt/nickel chokepoint' framing is DEMOTED to intensity decoupling: Cobalt Institute 2025 shows absolute EV cobalt demand still rising through 2024; 2025 cobalt price was moved by DRC export policy, not by LFP share; nickel weakness is Indonesian HPAL/NPI supply. ENGINE lead claim:pantry:1ffa05b1c097 was the question, not evidence.

★ For the kids: More cars use iron batteries with no cobalt, but so many cars are being built that the world still uses more cobalt than before.

Technical: Olivine LiFePO4, 3.2-3.4 V, ~160 mAh/g; kg Co per GWh fell but fleet GWh grew faster; LMFP tracked separately.

Frozen prediction: LFP+LMFP are >=50% of global EV battery GWh in calendar 2026 per IEA Global EV Outlook 2027.

HELD modeled across all lenses. 2026 evidence is pilot: QuantumScape Eagle Line inaugurated 2026-02-04 after B1 QSE-5 samples (Oct 2025); Samsung SDI Suwon pilot, H2 2027 target; Idemitsu sulfide-electrolyte pilot plant construction started 2026-01-29 (several hundred t/yr, mid-2027); Toyota 2027-28. Zero all-solid-state cells in customer cars; dates have only moved right.

★ For the kids: Solid batteries are being built on small test lines by several big companies; the first cars with them are promised for 2027-28, not yet.

Technical: Argyrodite/LGPS-class sulfides; bottlenecks are Li2S precursor supply, stack pressure and dry-room H2S handling; oxide ceramic separator (QS) is the main alternative.

Frozen prediction: Through 2027-12-31 no automaker delivers >=1,000 customer vehicles on all-solid-state sulfide or oxide-separator cells (p=0.85); by 2028-06-30 no Toyota/Lexus all-solid-state vehicle reaches a retail customer (p=0.7); Idemitsu reports pilot-plant completion by 2027-09-30 (p=0.6).

HELD modeled across lenses. Sila Moses Lake operating from autumn 2025 (~2 GWh-equivalent); Group14 BAM-2 ~90% built with 2026 furloughs; $300 M round and DOE offer are funding events (excluded). No Mercedes G 580 Titan Silicon customer deliveries confirmed; EV-dominant silicon is not in a customer car.

★ For the kids: Silicon can hold far more charge than graphite; the factories are built, but the first cars using it have not reached buyers yet.

Technical: Nano-Si in porous carbon scaffold; EV gates are calendar life and swelling at >30-50 wt% Si, and cost vs graphite; no multi-year data.

Frozen prediction: By 2027-06-30 Mercedes-Benz confirms customer deliveries of a production vehicle with Sila Titan Silicon cells (p=0.45); by 2027-06-30 Group14 does not report customer-qualified commercial production from Moses Lake (p=0.6).

HELD open. Best bulk result is P-assisted L10 ordering in cast Fe-Ni (Ivanov et al., Advanced Science 2022) and the NITE powder route (Goto et al., Sci. Rep. 2017); neither reports bulk (BH)max above La-Co ferrite; no third-party magnet-grade measurement. NdFeB-replacement framing retired.

★ For the kids: Scientists are trying to copy a meteorite magnet using only iron and nickel. The atoms line up a little, but it is still weaker than a fridge magnet.

Technical: Target Ku ~1.3 MJ/m3, Ms ~1.3 T (Lewis et al. 2014); measured bulk Hc <0.1 T class. Promotion to modeled: independent bulk Hc >=0.3 T AND (BH)max >=5 MGOe.

Frozen prediction: P=0.85 that no bulk L10-FeNi sample with (BH)max >=5 MGOe is published in a peer-reviewed journal by 2027-12-31.

HELD modeled (all three lenses). Eramet H1 2026 (2026-07-29): Centenario 8,440 t LCE in H1 (70% of half-nameplate), Q2 average 85%, June 90%, FY guidance 17-20 kt vs 24 kt/yr - the 07-31 kill dle-ninety-percent-nameplate stands (an exit rate is not a year). The standing prediction ('no new-entrant DLE plant at >=80% for two consecutive quarters by 2027-12-31') is declared AT RISK: Q3 2026 (late Oct) grades it; if it fails, promote ONLY the adsorption-on-Mg-rich-continental-brine branch per dle-blanket-viability. CTR/Plum IV 'proven at commercial scale' (SEC EX-99.1, corpus verified) is footno…

★ For the kids: A factory in Argentina pulls lithium out of salty water with a special sponge and is almost at full speed; a Californian company calls its small test plant 'commercial', but its own footnote says one-fifteenth size.

Technical: Al-hydroxide sorbent adsorption, water elution, concentration to carbonate; grading metric = quarterly output / nameplate from audited releases, never issuer 'commercial scale' language.

Frozen prediction: Standing: by 2027-12-31 no new-entrant DLE plant (commissioned 2023+) reports two consecutive quarters >=80% of nameplate (author P(holds)=0.3, watch on Eramet Q3 2026); Eramet FY2026 Centenario lands inside 17-20 kt LCE (p=0.65); CTR/ACR announces Stage-1 lithium FID by 2028-03-31 (p=0.35).

TIER HELD at viable, with the evidence base correctly restated rather than quietly preserved — this is the right way to save a tier. Li-ion hydrometallurgical recycling IS commercially shipping at scale, overwhelmingly in China, where the IEA reports two-thirds of all global battery-recycling capacity growth since 2020. The original founding justification ('commercial plants are operating: Redwood Materials, Li-Cycle') is retired: Li-Cycle filed bi-national insolvency in May 2025 with the $485M Rochester Hub unfinished and was sold to Glencore for $40M (closed 2025-08-08); Redwood cut jobs an…

★ For the kids: Old batteries really are mined for their metals — mostly in China, and by one big American company. Another American company everyone pointed at went broke before its plant ever opened.

Technical: Harper et al., Nature 575, 75-86 (2019) for route taxonomy. Redwood's '100 GWh CAM' is a target, not output, and Redwood publishes no CAM tonnage — which made the standing CY2026 >=50 GWh prediction ungradeable as written and forced the restatement below. Standing rule added to the tier-support checklist after the Natron/PBA and Li-Cycle precedents: cross-check every company named as tier support against insolvency filings before the tier is re-affirmed.

Frozen prediction: No US or EU recycler reports 50,000 t/yr or more of ACTUAL black-mass throughput derived from end-of-life EV packs (as distinct from production scrap) in any calendar year before end-2028.

HELD open; Lyten's Northvolt asset purchases (2025-10-15, 2026-02-26, 2026-03-13) are not Li-S maturity signals - H2-2026 Skelleftea shipments are Northvolt NMC cells; Li-S remains pilot in San Jose.

★ For the kids: Sulfur batteries could be light and cheap; one company bought a huge factory to build them, but the batteries are still test pieces.

Technical: Polysulfide shuttle and Li-metal anode life remain the gates; Lyten uses 3D-graphene sulfur cathodes.

Frozen prediction: Through 2027-06-30 no Li-S BESS >=1 MWh is commissioned for a paying customer. p=0.80.

HELD open. GM/LGES target LMR prismatic pre-production late 2027, commercial 2028 (GM newsroom 2025-05-13); Fan et al. (Nature 2026) lab-scale fast-formation mitigation of voltage decay. Nothing shipped; voltage fade and O-loss remain the commercial gate.

★ For the kids: Batteries with lots of cheap manganese would cost less but slowly lose voltage; GM plans to sell them in 2028.

Technical: Li-rich layered-layered xLi2MnO3.(1-x)LiMO2 with anionic redox; residual-Li self-pinning is a manufacturing-side mitigation orthogonal to doping/coating.

Frozen prediction: LGES publicly confirms start of LMR prismatic pre-production by 2027-12-31 (p=0.50); by 2028-12-31 no LMR prismatic cell is in a customer-delivered GM vehicle (p=0.6).

Graphite barely intercalates sodium, so Na-ion cells use hard (non-graphitizable) carbon anodes — genuinely shipping inside certified commercial Na-ion cells (CATL Naxtra, certified GB 38031-2025, 2025; also HiNa, Natron), which clears the literal commercial-shipping bar for 'viable'. SKEPTIC CAVEAT: a second proposer rated the same anode 'modeled'; I retain 'viable' on the shipping evidence (not market maturity) but flag that Na-ion is still a <1% niche and capacity (~300-350 mAh/g), first-cycle efficiency, low-voltage-plateau safety and biomass/pitch precursor variability remain open optimi…

★ For the kids: Sodium is too fat to fit the usual pencil-lead anode, so salt batteries use a special sponge-like carbon instead — and it already works in real cells.

Technical: Stevens & Dahn, J. Electrochem. Soc. 147, 1271-1273 (2000); Saurel et al., Adv. Energy Mater. 8, 1703268 (2018). Shipping in commercial Na-ion cells; precursor supply and first-cycle efficiency remain open.

Frozen prediction: Through end of 2028 hard carbon remains the dominant commercial Na-ion anode, no alloy/conversion anode displaces it in a shipped cell, and no maker cites hard-carbon precursor supply as a binding production constraint.

HELD open. Liu et al. (Nature 2026-03-17): 2.7 Ah anode-free pouch, 508 Wh/kg, 100 cycles at 100% DOD / 250 at 80% DOD to 80% retention - single-lab, short life. Nature Reviews Clean Technology 2025 reviews scalability limits.

★ For the kids: A battery with no anode at all is lighter but wears out fast; a 2026 result stretched it to 250 gentle cycles.

Technical: Zero excess Li; life set by SEI homogeneity and Li inventory loss (CE >99.9% needed for 500+ cycles).

Frozen prediction: Through 2027-12-31 no peer-reviewed anode-free pouch >=2 Ah reports >=500 cycles to 80% at 100% DOD with stack pressure <=1 MPa. p=0.75.

SKEPTIC ENDORSES the self-demotion modeled -> open. No primary evidence of pilot-scale cell production using an LHCE; 2025 reviews (Energy Storage Materials; ACS Applied Energy Materials) call practical application early-stage; TTE/BTFE diluents cost 10-100x carbonates. The prior modeled rating was plausibility, not a plant or shipping cell.

★ For the kids: A special salty electrolyte helps lithium metal behave, but the diluting liquid is very expensive and nobody makes it at factory scale yet.

Technical: LHCE = high local salt:solvent ratio diluted with a non-coordinating fluoroether; anion-derived inorganic SEI for Li metal and high-voltage cathodes at the cost of fluorine mass, cost and PFAS exposure.

Frozen prediction: Through 2027-06-30 no cell maker discloses a >=1 GWh/yr product line using a fluoroether-diluent LHCE. p=0.85.

MODELED, and now explicitly branch-tiered — 'redox flow' was three domains wearing one label. VANADIUM flow is commercially shipping (100 MW/400 MWh grid-connected in Dalian since 2022, multi-MW Sumitomo units running a decade) and would clear the viable bar alone, but carries its own vanadium chokepoint shared with NVPF sodium cells. IRON flow is pilot-scale with a financially distressed vendor set. ORGANIC (quinone/viologen) flow is lab-scale with unsolved molecular crossover and degradation. The aggregate is held at modeled by its weakest branches, and every frozen prediction must NAME its…

★ For the kids: Liquid batteries in tanks: the vanadium kind really works today, the iron kind is still being tested, and the plastic-molecule kind is still in the lab.

Technical: Skyllas-Kazacos et al., J. Electrochem. Soc. 158, R55-R79 (2011); Darling, Gallagher, Kowalski, Ha & Brushett, Energy & Environmental Science 7, 3459-3477 (2014) shows the abundant-material (iron/organic) branches only beat vanadium when active-material cost dominates — i.e. the branch the chokepoint-relief thesis needs is exactly the branch with the weakest evidence.

Frozen prediction: By 2028-12-31 no all-organic aqueous flow battery reaches >=10 MWh of commissioned commercial capacity, while vanadium flow exceeds 5 GWh cumulative commissioned.

VIABLE reaffirmed and correctly tiered — GBD is replication-grade, multi-producer and in volume production, cutting Dy/Tb usage by up to ~70% at equivalent coercivity by placing heavy rare earth only in the grain-boundary shell. Two extensions this round: spatially selective diffusion that hardens only the corners/edges of a rotor magnet that actually see demagnetizing fields, and Proterial's announced fully HRE-FREE sintered grade (Br 1.42 T, HcJ >=1830 kA/m). The caveat is binding: that GRADE is an announcement with lab-measured properties and April-2026 sample readiness, so it sits at mode…

★ For the kids: Strong magnets need a rare, expensive ingredient to survive heat. Engineers found you only need it on the skin of each tiny grain, like salting the crust of bread instead of the whole loaf.

Technical: Dy/Tb substitutes for Nd in a thin Nd2Fe14B shell, raising the local anisotropy field at the reverse-domain nucleation site without diluting bulk magnetization — which is why Br and (BH)max survive while HcJ rises. Frontier: HRE-lean diffusion sources (Dy-Ce-Cu), FEA-driven spatially selective diffusion (Durgun et al., Advanced Engineering Materials, 2025), and HRE-free grades via grain refinement. Asymmetry that matters for the wedge: GBD destroys demand for Tb/Dy specifically while leaving NdPr demand intact or growing.

Frozen prediction: Heavy-rare-earth (Tb+Dy) content per kW of traction magnet in teardown-assayed 2028 model-year vehicles falls at least 30% below the 2024 model-year baseline, while NdPr content per kW does not fall.

Viable holds cleanly: BMW Gen-5/Gen-6 EESM in mass production (millions of units, <1% highway efficiency gap to PM motors), Renault wound-rotor motors shipping for over a decade, Nissan/VW/ZF programs following — multiple OEMs, replication-grade commercial shipping. The demand-side relief valve on the NdPr chokepoint.

Technical: BMW Gen-5/Gen-6 EESM; Renault wound-rotor lineage; Nissan/VW/ZF programs.

Frozen prediction: >=3 distinct OEM platforms in mass production with magnet-free primary-drive motors by 2027-12-31; RE-free/externally-excited motors exceed 10% of new global BEV traction-motor units by end 2028 while PMSM stays >65%.

MODELED held, with a raised internal prior. Ce-substituted NdFeB (dual-main-phase and Ce-rich single-phase routes) is produced at commercial volume in China for non-traction grades while the high-performance end remains lab-validated — and the strongest new evidence for the latter is a single unreplicated ML-designed magnet, so it raises a prior rather than a tier. This is the highest-leverage substitution in the magnet lens for a supply reason: cerium and lanthanum are the OVERSUPPLIED co-products of every rare-earth mine, so Ce/La loading simultaneously relieves NdPr demand and monetizes a …

★ For the kids: When miners dig up rare earths they get a mix — the cheap parts come out whether anyone wants them or not. Building good magnets from the cheap leftovers fixes two problems at once.

Technical: Ce2Fe14B has lower anisotropy field and lower saturation magnetization than Nd2Fe14B, and Ce's mixed valence (Ce3+/Ce4+) makes substitution non-linear in composition; the dual-main-phase route co-sinters separate Ce-rich and Nd-rich powders so the Ce-rich phase carries volume while the Nd-rich phase carries coercivity. With NdPr oxide roughly doubling from ~$49/kg (Dec 2024) to ~$104/kg (Mar 2026), the incentive is the strongest in a decade. Promotion needs a named commercial grade at published spec from a non-Chinese producer, not another lab magnet.

Frozen prediction: No non-Chinese producer announces a commercial Ce-substituted NdFeB grade with published Br/HcJ spec before 2027-12-31.

HELD modeled, split maintained (bonded SmFeN shipping from Nichia/Sumitomo Metal Mining; sintered remains a lab compact - Saito & Nishio-Hamane, JMMM 605, 172320, 2024, SPS-DC at 473 K, 9.45 kOe, no (BH)max datasheet). Samarium's April-2025 export-control status caps the family's supply-security value.

★ For the kids: A samarium magnet powder that must be pressed cold because heat destroys it; no factory sells the solid version yet.

Technical: Sm2Fe17N3 decomposes above ~600 C; sintered promotion bar: producer datasheet (BH)max >=30 MGOe and HcJ >=1 T.

Frozen prediction: P=0.7 that no sintered SmFeN magnet with datasheet (BH)max >=30 MGOe is commercially offered by 2028-12-31.

Hard ferrites (SrFe12O19 / BaFe12O19) are the highest-volume permanent magnet by tonnage and are fully rare-earth-free; La-Co doping raises coercivity/energy product. They are the incumbent rare-earth substitute, but their energy product (~30-40 kJ/m3) is ~a tenth of sintered NdFeB, confining them to low-torque-density uses and larger, heavier motors. Skeptic verdict: genuinely 'viable' (massive commercial scale) but performance-capped — not a high-grade NdFeB replacement.

★ For the kids: The most common magnets are cheap ceramic ones with no rare metals — they just aren't strong enough for small, powerful motors, so engineers use them where size and weight matter less.

Technical: Pullar, Progress in Materials Science 57, 1191-1334 (2012). Commercially dominant by volume; performance-capped.

Frozen prediction: Ferrite will remain the dominant rare-earth-free magnet by tonnage through 2028 but will not displace NdFeB in high-power-density traction motors.

HELD at modeled. HyProMag HPMS units commissioned 2024-25; Noveon uses recycled feed; USAR reports pilot-scale swarf flowsheet testing (8-K 2026). Combined ex-China recycled NdFeB output is hundreds of t/yr; end-of-life collection is the binding constraint.

★ For the kids: Cutting magnets into shape wastes up to a third of the metal as grindings. New plants turn those grindings back into magnets.

Technical: HPMS demagnetises scrap at ~1-4 bar H2; swarf (15-30% of magnet mass) is the cleanest feed; oxidised swarf needs hydromet.

Frozen prediction: Ex-China recycled-feed sintered NdFeB output does not exceed 2,000 t in calendar 2027.

DEMOTION ENDORSED, modeled -> open. The modeled tier rested on a commercial vehicle carrying Hydro-to-Cathode direct-precursor recycling to scale; that vehicle failed. Ascend Elements filed Chapter 11 on 2026-04-09 (S.D. Tex.) after a structural capacity-versus-feedstock mismatch, with a $164M DOE cathode grant mutually cancelled in early 2025 and a further $316M cancelled thereafter, and its Kentucky plant halted when an offtaker withdrew. There is no operating commercial direct-cathode regeneration line in the US or EU. The relithiation chemistry is unchanged and still promising and Chinese…

★ For the kids: Instead of melting a worn-out battery down to raw metal, this gently refreshes the old part so it works like new again.

Technical: Shi, Chen & Chen, Green Chemistry 20, 851-862 (2018) (hydrothermal relithiation of LiCoO2); DOE ReCell pilot lines. Demonstrations largely lab/coin-cell.

Frozen prediction: No US or EU facility operates a direct cathode regeneration line producing more than 5,000 t/yr of qualified cathode active material before end-2029.

Bioleaching uses acidophilic bacteria/archaea (e.g. Acidithiobacillus) to dissolve metals at ambient conditions without smelting. It is commercially MATURE for copper (~15-20% of world output) and refractory-gold pretreatment, but recovery of battery metals (Li/Co/Ni from black mass) and critical metals from e-waste is lab/early-pilot: slow kinetics, low pulp density, microbial toxicity cap throughput. Skeptic verdict: the wedge-relevant application is 'open', even though copper bioleaching itself is viable.

★ For the kids: Tiny microbes eat their way through rock and old electronics to free the metals, no giant furnace needed — but only copper is done at scale so far.

Technical: Johnson, D.B., Current Opinion in Biotechnology 30, 24-31 (2014): commercial for copper/gold, emerging for battery metals/e-waste.

Frozen prediction: No commercial-scale bioleaching plant processing spent-battery black mass or e-waste for battery metals (>1,000 t/yr) will be operating before end of 2028.

HELD at open. Lanmodulin dimer separation (Mattocks et al., Nature 2023; Dong et al., ACS Cent. Sci. 2021) is bench-scale; no pilot has processed tonnes. Binnemans & Jones (J. Sustain. Metall. 2023) explain why non-SX routes lose on reagent cost at scale.

★ For the kids: Some bacteria make a protein that grabs only certain rare-earth metals. Scientists are trying to use it as a sorting tool, but it only works in test tubes so far.

Technical: Protein immobilisation capacity ~0.1-0.3 mmol/g, acid-eluent cycle stability and Fe/Al/Th pre-treatment are the unsolved scale-up variables.

Frozen prediction: No non-SX REE separation route reports a continuously operated pilot producing >=1 t/yr of >=99% single-REE oxide by 2027-12-31.

HELD open. Korea Zinc's $6.6bn Clarksville smelter (announced 2025-12-15; site prep 2026, commercial 2029, full 2030) is the most credible ex-China multi-minor-metal flowsheet and delivers nothing before 2029; China's Ga/Ge/Sb US-export ban suspended only to 2026-11-27.

★ For the kids: Tiny metals like gallium ride along inside zinc ore. A Korean company is building a huge plant in Tennessee to catch them - not until 2029.

Technical: Ga/Ge/In/Te report to zinc-refinery residues and copper anode slimes; recovery economics depend on host-smelter throughput.

Frozen prediction: Standing: neither Clarksville nor Rio Tinto Vaudreuil reaches commercial Ga or Ge production before 2026-12-31; NEW: Clarksville does not produce saleable Ga or Ge before 2029-12-31 (P=0.7).

HELD at open. Three years after MOFCOM No. 23, ex-China primary gallium is pilot/announcement stage (Rio Tinto/Indium Corp Vaudreuil ~3.5 t/yr demo; Metlen; Ramaco PEA); USGS MCS 2025 lists no US primary production and China ~98% of primary low-purity gallium. Verdict on the engine's question: controls accelerated announcements and stockpiling, not delivered capacity. byproduct-minor-metal-recovery (Ga/Ge/Sc/Sb) folded into this record.

★ For the kids: Almost all gallium is a leftover from making aluminium in China. Other countries are planning to collect their leftovers too, but the plans are still plans.

Technical: Bayer-liquor by-product at ~50-100 ppm Ga via ion-exchange; recovery units are sized to host refinery throughput and cannot scale independently; same for Ge (zinc residues) and Sc (Ti/Ni wastes).

Frozen prediction: No ex-China primary gallium facility >=10 t/yr is commissioned by 2027-12-31 and USGS MCS 2028 reports US primary gallium as nil/withheld.

Correctly open: iridium supply (~7-8 t/yr) is a hard bottleneck for multi-GW PEM buildout; lab thrifting and Ir-free anode results are not in any commercial stack at reduced loading, and weak hydrogen demand has deferred the crunch.

Technical: Minke et al., Int. J. Hydrogen Energy (2021); DOE H2NEW targets.

Frozen prediction: No major PEM OEM ships a commercial stack at <=0.1 g-Ir/kW (or Ir-free anode) by 2028-12-31; a Ru-majority/heavily-thrifted anode may ship, but no fully iridium-free PEM at >1 MW.

SKEPTIC: PROMOTION REFUSED, held at modeled. The battery lens promoted to viable on a 'doctrine 6' single-producer criterion that does not exist in data/doctrine/materials-review.json; the live feed record (asOf 2026-08-26) states the promotion criterion as 'a second maker ships dry-cathode cells OR an independent teardown confirms Tesla's'. Tesla's Q4/FY2025 update letter (2026-01-28) stating dry anode+cathode 4680 in Model Y packs is the same self-attestation class the 2026-07-31 skeptic kill refused (single manufacturer, single form factor, no teardown, yield and volume share undisclosed).…

★ For the kids: Instead of painting battery powder on wet and baking it, factories press it on dry. Tesla says it works for both sides now, but nobody outside Tesla has checked, so we keep the 'promising' label.

Technical: Dry cathode was the hard step (fragile, reactive cathode powders crack on calendering); Tesla's Nov 2025 patent describes a Gen-2 fibrillation process. Cost claim (LGES): 17-30% electrode-manufacturing cost reduction by removing NMP recovery and drying ovens. Unverified: Tesla's yield and volume share of dry-cathode cells.

Frozen prediction: By 2027-06-30 a second manufacturer (not Tesla) states in a primary document that dry-processed cathodes are in cells shipped to paying customers. p=0.30.

Genuinely commercial with multi-maker replication: Chinese cathode makers have shipped single-crystal NMC since ~2018 and LG Chem began mass production in 2023. Grain-boundary-free particles resist intergranular cracking, enabling several-thousand-cycle 'million-mile' lives; chokepoint relevance is dematerialization-by-longevity.

★ For the kids: Make each battery grain one perfect crystal instead of a crumbly clump, so it can last as long as the whole car.

Technical: Harlow et al., J. Electrochem. Soc. 166, A3031 (2019); National Science Review 10, nwad252 (2023); LG Chem single-crystal high-nickel mass production (Cheongju, 2023).

Frozen prediction: By end of 2028 single-crystal material exceeds 30% of global high-nickel (>=80% Ni) cathode shipments.

NEW split out of the lumped silicon/lithium-metal domain: Si-C anodes ship at scale in 2025-26 flagship phones from Honor, Xiaomi, Oppo, Vivo, OnePlus (Honor Magic V6 at 921 Wh/L, ~25% Si). Multi-vendor commercial shipping = viable. Lumping this with lithium-metal was distorting the ledger in both directions.

★ For the kids: The newest phones already use sand-cousin silicon in their batteries — that part is in your pocket, not a lab dream.

Technical: Nano-Si in carbon matrices tolerates swelling at 15-25% Si in small-format cells with sophisticated charge control; automotive duty cycles remain the unproven frontier and stay in the EV domain at modeled.

Skeptic demotion modeled -> open. LiNi0.5Mn1.5O4 offers a ~4.7 V plateau with no cobalt and little nickel, but after 15-20+ years its binding failures — carbonate electrolyte oxidation above ~4.5 V and Mn dissolution poisoning the anode — remain unsolved outside the lab, and nobody mass-produces it. A persistent unsolved failure mode with zero commercial cell is 'open', not 'modeled'.

★ For the kids: A battery recipe that works in the lab but slowly burns up its own liquid at full power — it has been 'two years away' for fifteen years.

Technical: Manthiram, Chemelewski & Lee, Energy Environ. Sci. 7, 1339 (2014). Honest upgrade trigger: LHCE/fluorinated-electrolyte pairings escaping pilot scale.

Frozen prediction: No LNMO-cathode EV cell reaches mass production (>100k vehicles/yr) before end of 2029.

Skeptic base-rate anchor. A rechargeable Mg prototype (Aurbach, Nature 2000) and an 'ultrafast' Al-ion cell (Nature 2015) are 25 and 11 years old with zero commercial products. Multivalent ions promise 2-3x charge per ion but pay with sluggish solid-state diffusion and electrolyte incompatibility. Any chokepoint-relief story leaning on multivalent chemistry inherits this base rate.

★ For the kids: Batteries whose particles carry two or three charges each sound like a cheat code, but after 25 years none has left the lab.

Technical: The 2015 Al-ion cell stored charge by AlCl4- intercalation into graphite at ~2 V — capacitor-adjacent. Divalent Mg2+ mobility in oxide hosts is the core unsolved physics.

Frozen prediction: No rechargeable magnesium, calcium, or aluminium battery is commercially shipping in any product by end of 2030.

Echion XNO niobium-oxide anode reached volume material production — CBMM inaugurated a 2,000 tpa (~1 GWh/yr) plant in Araxa (Nov 2024) — and commercial cells launched with GUS at Battery Japan 2026 (80% in 5 min, -30C charging, >10,000 cycles). Modeled: the material ships, but fleet-scale end-use is only beginning. Promote on documented multi-customer fleet operation.

★ For the kids: A different metal (niobium) lets batteries gulp a charge in five minutes even in freezing cold — trucks and buses get it first.

Technical: Nb-based Wadsley-Roth oxides intercalate Li fast at ~1.6 V vs Li, trading energy density for power, cycle life, and low-temperature capability — a niche complement to graphite, not a replacement.

SKEPTIC: OPEN, binding on all lenses. The battery lens's 'KEEP modeled' is overruled by the 2026-07-31 kill iron-air-modeled-tier (present in the feed ledger) and by the three other lenses that restore open. All 2026 evidence is company-origin: Form Energy's Series G release (2026-08-12; 80 GWh backlog; self-reported >100 h discharge at Cambridge MN in Q4 2025) and the Great River Energy 1.5 MW / 150 MWh site still 'expected online during 2026'. No utility, ISO or third-party dispatch, round-trip-efficiency or availability record exists. Backlog, financing and self-reported milestones are exc…

★ For the kids: A battery that rusts iron on purpose to store power for days. One is being switched on in Minnesota, but the only report card so far was written by the company itself.

Technical: Fe/FeOOH reversible rusting in alkaline electrolyte, air cathode, ~100 h duration, ~40-50% round-trip efficiency class (company). Upgrade to modeled: an independent (utility/regulator) performance report from Cambridge; to viable: >=2 independent utility-owned sites with published dispatch data.

Frozen prediction: Great River Energy or Form Energy publicly confirms Cambridge in commercial operation by 2026-12-31 (p=0.55); by 2027-06-30 no public utility/ISO/Form-GRE document reports a single continuous discharge >=100 MWh from Cambridge (p=0.6); by 2027-12-31 an independent party publishes measured Cambridge RTE and availability (p=0.5).

OPEN, reaffirmed against a rising publicity gradient (an R&D 100 2025 award, 2025-2026 reviews). Best reported performance remains 9.1 MGOe (72.4 kJ/m3) in a BONDED magnet from surfactant-assisted ball milling of annealed melt-spun ribbons; phase stabilization, scalable production and bulk-form engineering are unresolved per the reviews themselves. The genuinely distinctive property is a POSITIVE temperature coefficient of coercivity up to ~250-300 C — the opposite of NdFeB — which is attractive exactly where GBD-hardened NdFeB is most expensive.

★ For the kids: A magnet that gets stronger as it heats up, which is very unusual. So far it only works well in small lab batches.

Technical: Low-temperature-phase MnBi forms through a peritectic reaction with narrow compositional tolerance, Mn segregation is severe, and the phase decomposes near 355 C while bismuth melts at 271 C — so conventional sintering is unavailable. Calibration for the 'gap magnet' framing: hard ferrite is ~4-5 MGOe, sintered NdFeB ~50 MGOe, and the commercially empty band is roughly 5-15 MGOe. MnBi at 9.1 MGOe is in that band on paper — but bonded NdFeB already occupies part of it, so passing the gap-magnet test requires cost per unit (BH)max below bonded NdFeB, not a (BH)max number.

Frozen prediction: No MnBi magnet reaches commercial production above 100 t/yr before 2028-12-31, and no peer-reviewed BULK MnBi magnet exceeds 15 MGOe with independent replication in that window.

Correctly open despite strengthening lab evidence: Cu-added anisotropic SmFe12 sintered magnets hit record ~1.5 T coercivity (Acta Mater. 2023) and the intergranular-phase mechanism is resolved (Acta Mater. 2025) — but there is no pilot line or commercial bulk magnet anywhere, and remanence is handicapped by nonmagnetic phase fraction. Best-physics candidate, zero scale-up evidence.

Technical: Cu-added SmFe12 sintered coercivity record (Acta Mater. 2023); intergranular mechanism (Acta Mater. 2025).

Frozen prediction: No production EV traction motor ships with ThMn12-type or tetrataenite magnets before 2028-06-30; no bulk 1:12 magnet with RT coercivity >=1.2 T AND (BH)max >=200 kJ/m3 is independently replicated before end of 2030.

Hot deformation builds aligned nanocrystalline NdFeB with traction-grade heat resistance and ZERO Dy/Tb. Daido/Honda shipped it in the FREED hybrid drive motor from 2016 with dedicated mass production at Nakatsugawa — commercial proof that total HRE elimination in a traction magnet ships. Volume stays modest vs GBD sintered magnets, and Proterial's 2025 HRE-free SINTERED grade is development-stage, so viable attaches to the shipping hot-deformed lineage only.

★ For the kids: Extra-tiny neat crystals build a car-motor magnet needing none of the two scarcest metals — driving real cars since 2016.

Technical: Honda/Daido joint announcement (Jul 2016); Daido Electronics mass production from Aug 2016; Proterial HRE-free sintered grade (Jul 2025, development-stage).

Frozen prediction: By end of 2028 HRE-free NdFeB is offered for traction by >=3 major producers yet stays <20% of traction-magnet volume; GBD remains the dominant Dy/Tb lever.

HELD open. Tesla's 2026-09-04 Cybercab drive unit is stated 'rare-earth-free' (Musk on X; Electrek 2026-09-04) with magnet chemistry UNDISCLOSED (ferrite inferred; alnico or magnet-less topology not excluded) and ~45 units registered. Count: 2 prototypes + 1 production program of undisclosed chemistry at tens of units. Not evidence for ferrite-only >=100 kW at volume.

★ For the kids: Tesla says its robotaxi motor has no rare-earth magnets, but it has not said what the magnets are made of, and only a few dozen cars exist so far.

Technical: Ferrite ~4-5 MGOe vs NdFeB 40-50 forces higher pole count/speed and larger rotor; La-Co ferrite + spoke topology recovers part of the gap. Falsifier source: independent teardown naming the magnet plus registration counts.

Frozen prediction: P=0.8 that no EV model selling >10,000 units/yr ships a ferrite-only PM traction motor >=100 kW by 2028-12-31; P=0.6 that an independent teardown identifies the Cybercab motor's magnet material by 2027-06-30, P=0.65 conditional that it is a hard ferrite.

High-pressure acid leaching of laterites — a 2000s graveyard (Ravensthorpe, Goro) — now runs at massive commercial scale in Indonesia via multiple Chinese-built plants (~469 kt Ni-in-MHP in 2025; Indonesia ~47% of global primary nickel; capacity heading toward ~862 kt/yr in 2026). Multiple independent operating plants = replication-grade. The clearest modern case of extraction tech repricing a critical-metal chokepoint DOWNWARD: class-1 nickel fell to ~$15-16k/t and the 'nickel scarcity' narrative is dead.

★ For the kids: A once-failed way of squeezing nickel from tropical dirt now works so well the world has too much nickel instead of too little.

Technical: Argus Media (2025); Mysteel MHP data (2025-2026); ING nickel-surplus commentary (2025). Lens lesson: process tech + patient capital + one permissive jurisdiction can flip scarcity to glut in ~5 years.

Frozen prediction: LME class-1 nickel does not sustain a 6-month average above $25,000/t at any point before end of 2028.

HELD at open. Ramaco Brook Mine (Wyoming) Fluor PEA July 2025 and 2025-26 8-Ks describe a pilot facility for REE + Ga/Ge/Sc from carbonaceous shale; grades in the hundreds of ppm TREO; pre-feasibility; no saleable separated oxide from any coal by-product operation.

★ For the kids: The dust and rock left over from coal mining has tiny amounts of rare earths. One company says it can collect them, but it is still at the test stage.

Technical: Acid leach of 300-1,000 ppm TREO material with high gangue consumption; economics hinge on Ga/Ge/Sc co-recovery and the host coal mine carrying mining cost.

Frozen prediction: Ramaco Brook Mine reports no sale of separated REO or Ga/Ge product before 2027-12-31.

HELD modeled. Lynas FY26 (2026-08-26): 13,089 t REO, 7,260 t NdPr; Dy 240 t/yr + Tb 50 t/yr disclosed capacity (290 t/yr), Q3 FY26 Dy+Tb 8 t; Sm oxide first March 2026. The frozen '>=300 t FY2027 Dy+Tb, else demote' clause sits above nameplate; the prediction grades as written, the demotion CLAUSE is retired and a >=150 t companion is frozen.

★ For the kids: An Australian company now makes the rare heavy magnet metals outside China, but only a few tonnes a quarter so far.

Technical: Heavy separation is SX on mixed heavy-RE carbonate from Mt Weld (light-dominated); ex-China Dy+Tb a few hundred t/yr vs ~10 kt/yr Chinese.

Frozen prediction: Standing (grades as written): Lynas FY2027 combined Dy+Tb >=300 t; companion: FY2027 Dy+Tb >=150 t (P=0.45).

TIER HELD at modeled; MERGED from two duplicate proposals. The standing read is intact: Syrah Vidalia produces active anode material ex-China but slow qualification and Syrah's 2025 restructuring prove production without bankable offtake is not commercial viability — the 'producing but unqualified' trap. Two amendments endorsed. (1) The policy risk is dated, not diffuse: artificial-graphite anode materials, equipment and technology were placed under control by Announcement No. 58/No. 61 effective 2025-11-08 and suspended 2025-11-07/09 for twelve months to 2026-11-10, giving one dated test wit…

Technical: Syrah Vidalia AAM production; 2025 restructuring disclosures.

Frozen prediction: Through end-2028 China retains >80% of global artificial-anode-material production; Vidalia reaching >=50% nameplate for two consecutive quarters by 2027-12-31 is scouted AGAINST at 40% FOR; no ex-China AAM plant discloses in an audited filing that a majority of its needle-coke or pitch precursor was sourced outside China and Japan before 2028-12-31.

TIER HELD at modeled; the domain's own governing sentence is retired. This record said lithium conversion is 'a capacity/economics chokepoint, not a knowledge chokepoint' because 'the chemistry is a century old'. On 2025-07-15 MOFCOM and MOST added five lithium control points to the restricted technology catalogue: spodumene-to-carbonate/hydroxide extraction, metallic lithium and lithium-material preparation, brine extraction, and purification of lithium-bearing liquids. The controlled object is not the century-old chemistry but the modern cost-competitive Chinese flowsheet — precisely what a…

★ For the kids: Digging up lithium rock happens in many countries, but turning rock into battery powder still mostly happens in one — copying that step is proving expensive.

Technical: IEA Global Critical Minerals Outlook (2024, 2025); Albemarle Kemerton curtailments (2024). Maturity reflects ex-China delivery risk, not process risk.

Frozen prediction: By end-2028 China still converts over 55% of world lithium chemicals; at least one further Western conversion plant is curtailed, sold or cancelled; and no Western converter publicly discloses receipt of a MOFCOM technology-export licence for a restricted lithium process before 2027-12-31.

VIABLE, confirmed. CATL Shenxing (4C) shipping since 2023-24; BYD Super e-Platform (~1 MW / 10C-class) in Han L / Tang L production vehicles. A shipping commodity feature achieved by electrolyte-transport + electrode-architecture engineering, not new chemistry -- the honest headwind that compresses the differentiation window every premium chemistry (solid-state, silicon-dominant) counts on.

★ For the kids: Cheap iron batteries now charge in minutes in cars you can actually buy — which makes life much harder for the fancy future batteries.

Technical: CATL Shenxing (4C) shipping since 2023-24; BYD Super e-Platform (~1 MW, 10C-class) in Han L / Tang L production vehicles. Achieved via electrolyte-transport and electrode-architecture engineering, not new chemistry; compresses the differentiation window for solid-state and silicon-dominant premium chemistries.

DEMOTED viable -> modeled. Two of the four proposals wanted to keep this at viable on years of Bluebus/eCitaro service; one demoted it on the single-producer rule. The demotion wins, and consistency is the reason: the 2026-07-16 audit demoted DLE (one plant) and dry-electrode (one manufacturer) on exactly this bar. Blue Solutions is by its own statement the only company mass-producing LMP, ~1 GWh cumulative since 2011 (under one day of world Li-ion output), one site, one 60-80 C thermal regime, gen4 gigafactory not before 2029 — plus RATP's 2022 withdrawal of 149 Bluebus 5SE after two fires. …

★ For the kids: One French company makes these hot-running solid batteries for buses. One company doing one thing is not proof the world can do it — and Paris parked 149 of those buses after two caught fire.

Technical: Armand & Tarascon, Nature 451, 652-657 (2008) for the polymer-electrolyte lineage; Blue Solutions disclosures (2026): >1 GWh cumulative since 2011, sole mass producer, gen4 room-temperature cell targeted at a first gigafactory in 2029. The framing correction survives the demotion — LMP does qualify as all-solid-state under China's <5 wt% residual-liquid criterion effective 2026-07-01 — but that correction now rests on a MODELED domain and must be quoted with the tier attached, and with the 60-80 C operating qualifier.

Frozen prediction: No second independent producer mass-produces LMP cells before 2028-12-31, AND Blue Solutions' gen4 gigafactory does not begin cell production before 2029-01-01 with LMP output remaining single-site through 2028-12-31.

Viable holds: multiple Chinese producers (Tinci, ChunBo and peers) operate multi-ten-kiloton capacity with LiFSI blended into mainstream fast-charge electrolytes — commodity-scale shipping across several independent producers, which clears the replication bar. China-concentrated; adds a fluorochemical link to the chain (see pfas-fluorine-battery-materials for the regulatory tail risk). Scope boundary: this domain grades the SALT supply chain; formulation-class claims (LHCE/concentrated electrolytes) stay in concentrated-fluorinated-electrolytes at modeled.

Technical: Xu, Chem. Rev. 114, 11503-11618 (2014); LHCE lineage (Ren et al., Chem 2018); disclosed Chinese capacity expansions 2023-2025.

Frozen prediction: By end of 2028, LiFSI demand grows faster year-over-year than total lithium-ion GWh deployments in every year.

HELD modeled on the semi-solid branch only (WeLion 150 kWh packs in NIO vehicles since 2024); CATL 'condensed battery' 500 Wh/kg (April 2023) struck from tier support - no shipping cell, customer, teardown or filing three years on.

★ For the kids: Some new car batteries are like jelly instead of liquid; those are real. A super-battery shown on a stage three years ago has not turned up in any car.

Technical: Semi-solid = gelled/in-situ polymerised carbonate electrolyte with 5-15 wt% residual liquid; 'condensed' as marketed has no independent teardown.

Frozen prediction: By 2027-12-31 no third-party teardown or regulator-filed spec documents a series-production CATL 'condensed' cell at >=450 Wh/kg (P=0.8).

HELD modeled with a SKEPTIC FLAG: the new milestone (Idemitsu Chiba construction start 2026-01-29, ~USD 143 M, 1,000 t/yr Li2S, completion mid-2027; electrive 2026-02-02) is a construction start, which carries no tier weight under the standing rule that groundbreakings and funding are not evidence. The modeled tier rests on Idemitsu's existing small-scale Li2S pilot output, not on this announcement.

★ For the kids: To make solid batteries you first need lithium sulfide powder; Japan is building the first real plant for it, ready in 2027.

Technical: Li2S is the cost driver of argyrodite electrolytes; Idemitsu's route leverages refinery sulfur. 1,000 t/yr is pilot-to-early-commercial scale.

Frozen prediction: Idemitsu announces mechanical completion of the Chiba Li2S plant by 2027-09-30. p=0.60.

Correctly open: comprehensive chemistry (Prussian-blue cathodes, graphite anodes, ether electrolytes) but lab-to-prototype only; announced 18650 prototypes (Group1, 2024) have no independent replication and no commercial shipments.

Technical: Hosaka, Kubota, Hameed & Komaba, Chem. Rev. 120, 6358-6466 (2020).

Frozen prediction: No potassium-ion cell line reaches commercial production above 100 MWh/yr before end of 2029.

DEMOTION ACCEPTED modeled -> open. Eos FY2024 10-K: revenue ~$15.6M vs COGS ~$110M despite a $303.5M DOE LPO loan; Zinc8 creditor protection 2024; e-Zinc/Enerpoly pilot. Early-commercial-by-press-release. Re-promotion: a full fiscal year of positive gross margin at any Zn vendor, or >=500 MWh energized under third-party ownership.

Technical: Eos Znyth deployments + DOE LPO loan (Dec 2024); Redflow administration (Aug 2024).

Frozen prediction: Eos does not report positive annual gross margin for FY2026 (P=0.7).

HELD open under the standing intake filter. Niron's conditional $150M Department of War OSC loan (2026-08-07), Wood EPCM contract and 10,000 tpa 2028 expansion are financing/announcements, not a measured bulk (BH)max on a shipped part.

★ For the kids: A new iron-and-nitrogen magnet factory got a big government loan, but a loan is not a magnet measurement.

Technical: Alpha''-Fe16N2 metastable (>~200 C decomposition), low Ku; to modeled: OEM-disclosed pilot motor with independently measured magnet performance.

Frozen prediction: P=0.8 no production passenger-EV traction motor (>100 kW) using Fe16N2 ships before 2028-12-31; P=0.7 Niron publishes no third-party-measured (BH)max for a production part before Sartell's first commercial shipment (or 2027-12-31).

HELD modeled. Mid-2026 census: ~8,000-10,000 tpa announced nameplate (eVAC, Neo Narva, Noveon, USAR, MP Independence) vs zero disclosed shipped tonnage in any audited filing; MP Q2 2026 Magnetics revenue down 17% YoY to $16.5M with Independence in qualification; USAR Q2 2026 $5.8M revenue, no magnet tonnage (EDGAR EX-99.1 2026-08-10); Neo's 'one millionth magnet' is a piece count; Niron's conditional $150M OSC loan is financing and not NdFeB. EMAT's 'only DFARS-capable' claim (EX-99.1 2026-08-10, corpus verified) is retired.

★ For the kids: Lots of new magnet factories outside China are opening, but almost none has said how many tonnes it has actually made.

Technical: Chokepoint remains upstream (strip-cast alloy and metallization; LCM/USAR corpus lead 85ae217bdc95 verified as an operating UK alloy asset). Reading rule: nameplate, ramp targets, piece counts, MoUs and conditional loans carry no tier weight.

Frozen prediction: P=0.6 MP Materials reports annualized finished-NdFeB run-rate >=1,000 tpa by 2027-06-30; P=0.75 USAR cumulative sintered NdFeB shipments <500 t by 2027-06-30; P=0.7 no ex-China ex-Japan producer discloses >=1,000 t finished sintered NdFeB shipped in CY2026 in an audited filing.

HELD at open. TMC applied for NOAA DSHMRA permits (April 2025) after the US seabed-minerals executive order; no ISA exploitation code; processing only at pilot campaign scale (TMC/PAMCO 2024-25).

★ For the kids: Potato-sized rocks full of metal sit on the deep ocean floor. Companies want to vacuum them up, but the rules are not agreed and no one has done it for real.

Technical: Nodules ~1.3% Ni, 0.2% Co, 1.1% Cu, 27% Mn; manganese by-product credit dominates revenue and depends on a silico-manganese market that may not absorb it.

Frozen prediction: No company sells commercially processed Ni or Co product derived from seabed nodules before 2027-12-31.

Correctly open: bench cells hit battery-grade purity from extreme Mg/Li brines (Science 2024; Matter 2024) but all demonstrations are hours-to-100-h lab runs on the bench — earlier-stage than sorbent DLE (graded separately in direct-lithium-extraction), with electrode durability on fouling real brine unproven.

Technical: Li, Chen et al., Science 385, 1438-1444 (2024).

Frozen prediction: By end of 2028, no membrane-free/decoupled electrochemical lithium-extraction process operates at commercial brine scale (>5,000 t LCE/yr).

OPEN (demoted from modeled). Evidence is field trials and small demonstration harvests (Univ. Lorraine/Econick lineage; plots in Albania, Malaysia, Sabah) with no commercial nickel offtake, and the Indonesian HPAL nickel glut removed the economic headroom the modeled tier assumed. Yields ~50-120 kg Ni/ha/yr are marginal at post-2024 class-2 nickel prices without carbon/remediation co-payments.

★ For the kids: Some special plants drink up nickel from the soil, and you can harvest metal from them — but nickel got so cheap that farming it does not pay right now.

Technical: Field trials and demo harvests only (Univ. Lorraine/Econick lineage; Albania, Malaysia, Sabah), no commercial offtake. ~50-120 kg Ni/ha/yr is marginal at post-2024 class-2 nickel prices after the Indonesian HPAL glut, absent carbon/remediation co-payments. DEMOTED from modeled 2026-07-16.

TIER HELD at open — citation strength upgraded, tier unchanged, which is the honest outcome when new evidence corroborates rather than advances. The ~95% China share previously rested on USGS plus price-reporting-agency tier; the IEA's Global Critical Minerals Outlook 2025 now states the same figure independently and names HPMSM alongside purified phosphoric acid as an emerging chokepoint. That corroborates the FACT and changes nothing about the TIER, because the tier was never about the concentration number — it was about the ex-China buildout being pre-commercial and slipping (Element 25's …

Technical: USGS Mineral Commodity Summaries 2025 (Manganese); Benchmark/Fastmarkets HPMSM concentration data.

Frozen prediction: Through end-2028 China retains >85% of battery-grade HPMSM production; no ex-China plant reaches sustained >30% nameplate at battery grade.

Modeled holds: Almonty Sangdong completed Phase 1 commissioning (Mar 2026) and produced first saleable concentrate (Jun-Jul 2026) — first concentrate exists; steady-state shipping and downstream APT/carbide qualification do not. Watch the chokepoint moving one step down the chain to conversion capacity.

Technical: Almonty Sangdong Phase 1 commissioning (Mar 2026); first saleable concentrate (mid-2026).

Frozen prediction: Sangdong ships >=1,500 t tungsten concentrate in calendar 2027; falsified if reported 2027 shipments below that.

OPEN, and now datable. ECHA's Risk Assessment Committee adopted its FINAL opinion on the universal PFAS restriction on 2026-03-02, backing an EU-wide restriction with targeted derogations and keeping FLUOROPOLYMERS in scope on life-cycle grounds; the socio-economic committee's consultation closed 2026-05-25 (3,511 comments from 3,200+ organisations) with a final opinion due end-2026. This remains a RISK sensor, not a relief route: a broad restriction with narrow derogations TIGHTENS the PVDF and fluorinated-salt chokepoint. Fluorine-free binders and salts remain lab-scale.

★ For the kids: Europe is deciding which 'forever chemicals' to ban. Batteries are full of them — including the ones inside batteries advertised as safe.

Technical: Exposed materials: PVDF binder (Kureha, Arkema, Syensqo), LiPF6, LiFSI, FEC — and, less obviously, water-in-salt LiTFSI and fluoride-ion electrolytes, which are MORE fluorine-intensive per kWh than the incumbents they claim to displace. Score this domain jointly with fluorspar-hf-fluorochemical-supply: the two exposures point in opposite commercial directions and can offset, so netting them silently would hide both.

Frozen prediction: SEAC's final opinion (due 2026-12-31) retains fluoropolymers in scope with battery-related derogations of 13.5 years or less; separately, no mass-market cell (>1 GWh cumulative) ships fully PFAS-free before 2028-12-31.

HELD at open. CATL's second-generation Naxtra (April 2025) moved to layered oxide, away from the PBA chemistry of its 2021 first generation; Natron Energy (PBA, US) ceased operations in 2025. Peng et al. (Advanced Materials 2022) review lattice-water and vacancy limits.

★ For the kids: One sodium battery uses a blue pigment from old paintings. It turned out hard to make well, and the biggest company switched recipes.

Technical: Lattice water and Fe(CN)6 vacancies cap practical capacity near 120-140 mAh/g and drive gassing; drying protocols conflict with wet-precipitation economics.

OPEN. Chalmers demonstrated ~24-30 Wh/kg load-bearing multifunctional laminates with usable stiffness (Asp et al., 2021; 2024 follow-up). Two orders below pack-level Li-ion on energy; the honest metric is energy x stiffness where mass that must exist anyway becomes storage. Strictly lab-scale.

★ For the kids: Imagine the car's body panels secretly being the battery. It works in the lab, but holds far, far less energy than a real battery so far.

Technical: Chalmers (Asp et al., 2021; 2024 follow-up): ~24-30 Wh/kg load-bearing carbon-fibre laminates with usable stiffness — roughly two orders below pack-level Li-ion. Honest merit metric: energy x stiffness on mass that must exist anyway. Strictly lab-scale.

OPEN, tier unchanged, citation defect corrected. The prior record cited a bare 'Nature Energy 2023' with no authors — an evidence-quality failure for a domain resting on a single paper. The result is Kim, Wang, Xu et al., Nature Energy 8, 814-826 (2023): entropy-raised weakly-solvating electrolytes cut ion clustering, roughly doubling conductivity and cycling anode-free NMC622||Cu pouch cells at up to 2C. Coin- and small-pouch scale, mechanism still debated, no cell line publicly specifies a high-entropy formulation. A feeder into the fluorinated-electrolyte and sodium-ion tracks, not a stand…

★ For the kids: Mixing many different ingredients into the battery liquid instead of a few can make it work better — so far only in tiny test cells.

Technical: Kim, S.C., Wang, J., Xu, R. et al., Nature Energy 8, 814-826 (2023); Wang, Q. et al., Nature Communications 14 (2023), doi:10.1038/s41467-023-36075-1; high-entropy engineering also applied to NFPP thermal stability (ACS Nano, 2026). NEW STANDING RULE adopted: un-attributed journal-and-year strings are disallowed as evidence for any single-paper domain.

Frozen prediction: No commercial cell line publicly specifies a high-entropy electrolyte or high-entropy cathode formulation before 2028-12-31.

VIABLE attaches to the MAGNET TECHNOLOGY only: SmCo (SmCo5/Sm2Co17) is replication-grade, commercially shipping for five decades, the qualified standard for >250 C and defense (an F-35 carries ~23 kg). The feedstock is the open risk: China controls essentially all usable samarium production, Sm entered the April 2025 export controls, prices spiked and Western stockpiles are dwindling; ex-China Sm oxide is announcement-tier. SmCo has no NdFeB drop-in at high temperature -- the least-substitutable magnet chokepoint on the map.

★ For the kids: These fifty-year-old super-heatproof magnets guard jets and rockets — but nearly all of their special ingredient comes from one country.

Technical: SmCo5/Sm2Co17: replication-grade, five decades of commercial shipping, the qualified standard above 250 C and in defense (an F-35 carries ~23 kg). Feedstock risk: China controls essentially all usable Sm production; Sm entered the April 2025 export controls; prices spiked, Western stockpiles dwindling; ex-China Sm oxide is announcement-tier. No NdFeB drop-in at high temperature.

Frozen prediction: No ex-China facility produces >=100 t/yr separated samarium oxide by 2028-12-31, and no non-SmCo magnet qualifies into a >250 C US defense program before end of 2028.

OPEN. tau-phase MnAl(-C) has the cheapest raw-material basis of any hard-magnet candidate but the tau phase is metastable and decomposes during consolidation; best verified coercivities are powders/ribbons (~5.4 kOe Tb-doped, ~4.9 kOe milled). No bulk anisotropic magnet or catalog item after four decades; a 2026 review confirms bulk processing remains the open front. (Note: the record MnAl coercivity used terbium -- a rare earth quietly re-imported into a 'rare-earth-free' magnet.)

★ For the kids: A magnet made from two cheap common metals would be wonderful — but for forty years it keeps falling apart before anyone can make a solid one.

Technical: tau-MnAl(-C) is metastable and decomposes during consolidation; best verified coercivities are powders/ribbons (~5.4 kOe Tb-doped, ~4.9 kOe milled). No bulk anisotropic magnet or catalog item after four decades; 2026 review confirms bulk processing is the open front. The record coercivity used terbium — a rare earth re-imported into a 'rare-earth-free' magnet.

VIABLE material (NOES is mature, replication-grade, commercially shipping); the tracked question is a soft-magnetic chokepoint invisible to critical-mineral lists. Every rare-earth-relief route (EESM, ferrite, weaker RE-free magnets) raises high-grade thin-gauge NOES intensity per motor, while <=0.27 mm high-frequency capacity (Baosteel launched 0.10 mm in May 2025) is concentrated in a few Chinese/Japanese/Korean mills against demand forecast to climb ~8x (2020->2027). The bottleneck is rolling-mill capability, not element scarcity.

★ For the kids: Every trick for avoiding rare-earth magnets needs more of a special whisper-thin steel — and only a handful of mills in the world can roll it.

Technical: NOES is mature and replication-grade; the chokepoint is capability, not chemistry. <=0.27 mm high-frequency capacity (Baosteel launched 0.10 mm May 2025) concentrates in a few Chinese/Japanese/Korean mills; demand forecast ~8x 2020->2027. Every rare-earth-relief route (EESM, ferrite, weaker RE-free magnets) raises thin-gauge NOES intensity per motor.

Frozen prediction: Through 2028-12-31 no production EV (>10,000/yr) replaces NOES stator laminations with amorphous ribbon or SMC; thin-gauge NOES stays binding for every magnet-substitution route.

HELD modeled, scope NARROWED. Pantry lead 85ae217bdc95 verified in the local EDGAR rare-earth corpus: USA Rare Earth describes LCM (UK) as its strip-cast alloy source and is developing a 3,750 mtpa metal/alloy facility at LCM Europe plus 5,000 tpa strip-cast/metal/alloy at Stillwater. The evidenced ex-China capability today is ONE legacy strip-cast producer (LCM) at hundreds-of-tonnes scale changing hands - not new capacity; 'only proven ex-China producer' is an issuer uniqueness claim. Nothing ex-China approaches Chinese tens-of-kt strip-cast scale (Adamas 2024; USGS MCS 2025).

★ For the kids: Making a magnet is like baking: you need flour (the metal) but also someone who knows the dough recipe (the alloy). Outside China only one or two bakeries know it.

Technical: Induction melt -> Cu-wheel strip casting -> 0.2-0.4 mm flakes with columnar Nd2Fe14B grains; flake microstructure sets post-HD/jet-mill grain size; alloy know-how is the moat.

Frozen prediction: By 2027-12-31 no single ex-China producer publicly reports >5,000 t/yr actual (not nameplate) NdFeB strip-cast alloy output (P=0.75).

OPEN, entered as open not modeled. ReElement (Purdue ligand-assisted displacement chromatography IP) is commissioning modular containerized columns and claims first US ultra-pure REE production 'at scale' -- but 'at scale' is early-commercial small-volume, and throughput/cost-per-kg vs SX above a few hundred t/yr is unproven and largely self-attested. Independent third-party throughput/purity data is thin.

★ For the kids: A new filter-column trick makes super-pure rare earths inside shipping containers — but 'big production' so far really means small batches.

Technical: ReElement (Purdue ligand-assisted displacement chromatography IP) commissioning modular containerized columns; claimed first US ultra-pure REE production 'at scale' is early-commercial small-volume. Throughput and cost-per-kg vs SX above a few hundred t/yr unproven and largely self-attested; independent third-party data thin.

Frozen prediction: No chromatography-based plant reaches independently verified >=500 t/yr separated single-REE oxide before 2027-12-31.

OPEN (post-bankruptcy pilot-scale). Ambri's molten Ca-alloy / molten-salt / solid-Sb-cathode cell (MIT/Sadoway lineage) promises cheap long-cycle LDES from abundant materials, but Ambri filed Chapter 11 (2024) and sold assets to lenders; deployment is limited to pilots (Microsoft backup, Xcel Aurora, TerraScale) and 1 MWh containerized units, with a Perpetua Idaho antimony offtake tying it to the antimony chokepoint. Stays open until standardized systems operate at multiple grid sites through the restructured entity.

★ For the kids: A battery made of molten metal from a famous MIT lab went bankrupt before it could grow up — small test units still hum along at a few sites.

Technical: Ambri (MIT/Sadoway lineage): molten Ca-alloy / molten-salt / solid-Sb cathode. Chapter 11 in 2024, assets sold to lenders; deployments limited to pilots (Microsoft backup, Xcel Aurora, TerraScale) and 1 MWh containerized units. Perpetua Idaho Sb offtake ties the route to the antimony chokepoint.

Frozen prediction: No Ca-Sb liquid-metal product reaches >100 MWh cumulative commercial deployment before end of 2028.

DEMOTED viable -> modeled, and three duplicate proposals merged into this slug. The finding is real and important: LiPF6, LiFSI, FEC, fluorinated LHCE co-solvents, PVDF binder and the HF used to purify graphite all descend from acid-grade fluorspar (>97% CaF2), so three separately-tiered domains share one undeclared parent and chokepoint scores treating them as independent hedges were double-counting. But 'the chain is documented and concentrated' is a structural OBSERVATION, not an evidence tier for a technology, and the same crew concedes the ex-China acidspar buildout and the substitution …

★ For the kids: A white rock called fluorspar is quietly inside almost every battery — in the salt, the glue and the cleaning acid. Most of it comes from one country, and nobody was watching it.

Technical: USGS Mineral Commodity Summaries 2025/2026 (Fluorspar): China ~60-65% of mine output and a larger share of HF-to-LiPF6 conversion; fluorspar is a critical raw material in both the US and EU (CRMA) lists. Acid grade is ~57-58% of the market by application. Trade reports of a Chinese export-licence regime from January 2026 are price-reporting-agency tier and are NOT part of the tier claim. Two-sided risk: a PFAS restriction destroys demand for the same molecules a fluorspar squeeze makes scarce, so pfas-fluorine-battery-materials and this domain must be scored separately and can offset. Action:…

Frozen prediction: China's share of world fluorspar mine production remains >=50% in the USGS Mineral Commodity Summaries covering calendar 2028, AND no greenfield ex-China acid-grade fluorspar operation reaches >=100 kt/yr of acidspar concentrate before 2028-12-31.

DEMOTED modeled -> open, and a duplicate slug ('copper-sulfide-leaching') merged in. One proposal tiered this modeled on a commercial deployment since 2019 plus first copper at two sites; the other tiered it open because first metal is not a recovery curve. Open is correct: every recovery figure in the record (Jetti ~40-70% vs 15-30% conventional; Ceibo 70-80%; Nuton ~85%) is VENDOR-REPORTED, no independently audited full-cycle heap recovery at commercial tonnage exists anywhere, catalyst-decay economics only entered the open literature in 2025, and all deployments are brownfield assets with …

★ For the kids: Most copper is locked in a rock that refuses to dissolve, so it has to be melted in giant furnaces. Some companies say a helper chemical finally cracked it, but they have only made first small batches.

Technical: Passivation baseline: Li, Kawashima, Li, Chandra & Gerson, Advances in Colloid and Interface Science 197-198, 1-32 (2013). Catalyst decay is the economic variable: Krishnamoorthy, Ren, Chao & Dixon (UBC), SSRN preprint, ethylene thiourea as a model for Jetti reagents. Deployments: Jetti at Capstone Pinto Valley (commercial dump leach plus 50 t crib tests), Rio Tinto Nuton first copper at the Johnson Camp industrial demonstration (2025-26), Ceibo first copper in Chile (2026) plus a Glencore Lomas Bayas pilot. If this route ever works at greenfield scale it deflates a chokepoint the way HPAL de…

Frozen prediction: Through 2028-12-31 no chalcopyrite-leach operation publishes an independently audited copper recovery of >=70% over a full >=12-month heap cycle at >=25 kt/yr Cu; catalytically-leached primary chalcopyrite stays below 2% of global refined copper; and smelter-routed concentrate remains >85% of world primary copper.

DEMOTED modeled -> open. The proposal's own frozen prediction is that this result will NOT be independently replicated — a domain whose author expects non-replication cannot sit above the contested tier. The entire evidence base is one 2026 single-group paper (XGBoost + particle-swarm optimization, Ce-rich Nd-Fe-B at 40.1 MGOe / 12.7 kOe) with proprietary process windows, laboratory-button scale, and no independent measurement; the companion Ames roadmap is institutional, not experimental, and its exact article could not be verified. The distinction the crew drew (screened-only = open, closed…

★ For the kids: Instead of guessing which recipe makes a better magnet, computers now learn from old recipes and suggest new ones. One suggested recipe was actually made once, in one lab. Nobody has made a big one or checked it anywhere else.

Technical: J. Mater. Chem. A (RSC, 2026), doi:10.1039/d5ta10051h — ~98% remanence prediction accuracy, experimentally validated Ce-rich Nd-Fe-B, Nb-optimized HRE-free compositions claimed to pass the predicted Pareto frontier; author list not retrieved. Ames National Laboratory AI-driven permanent-magnet roadmap (2025-26) — organization and topic confirmed, article number NOT retrieved, so no volume/page may be quoted. The methodological read survives the demotion and is worth keeping: ML is more productive at optimizing a manufacturable system (Ce/La loading in NdFeB) than at conjuring a rare-earth-fre…

Frozen prediction: The ML-designed Ce-rich Nd-Fe-B result (40.1 MGOe, 12.7 kOe) is not independently replicated by a group with no shared authorship at >=1 kg magnet scale, reporting magnet mass and a full demagnetization curve, in peer-reviewed literature by 2027-12-31.

VIABLE HELD, scope made binding. Fe-Si-B amorphous ribbon and Fe-Si-B-Nb-Cu nanocrystalline ribbon (FINEMET, commercialised 1988) are replication-grade, multi-producer and shipping at scale in distribution-transformer cores, common-mode chokes and EV DC-DC / on-board-charger cores — that is commercial shipping by any reading. The tier covers THOSE USES ONLY. The aggressive claim inside the same record — nanocrystalline stator cores replacing non-oriented electrical steel in traction motors at 64-75% lower iron loss — is laboratory/prototype and is retired as a sub-thesis so it cannot inherit …

★ For the kids: Motor and transformer iron is usually a stack of thin metal pancakes. These ribbons are made by squirting liquid metal onto a freezing spinning wheel so fast the atoms never line up, which wastes far less energy — but the ribbon is as brittle as a dry noodle, which is why it is not in car motors.

Technical: Rapid solidification (~10^6 K/s) gives 15-25 um ribbon; controlled devitrification precipitates ~10-15 nm bcc Fe-Si grains below the ferromagnetic exchange length so magnetocrystalline anisotropy averages toward zero (Herzer random-anisotropy model). Permeability ~48,000 at 1 A/m / 100 kHz, core loss ~94 kW/m3 at 0.2 T / 100 kHz (Energies 18(3), 482, 2025; JMMM FINEMET review, 2016). Motor blockers: stacking factor, brittleness hostile to stamping, and Bs ~1.2-1.3 T against NOES ~1.9-2.0 T, so torque density falls even as loss falls. Contains zero rare earths and only modest cobalt — a genuin…

Frozen prediction: No series-production passenger-EV traction motor ships with an amorphous or nanocrystalline stator core before 2028-12-31; the tier remains bounded to transformer, choke and converter cores.

VIABLE held, with a binding single-vendor caveat attached. Molten sodium-sulfur and sodium-nickel-chloride cells behind a beta-alumina separator are the only non-lithium electrochemical family with a multi-GWh, multi-decade FIELD record — roughly 5 GWh across 200+ sites (NGK), plus a second producer for the ZEBRA branch. That is commercial shipping with two decades of replicated deployment, so the tier survives the single-producer bar that demoted LMP — but only because the deployment count and operating years differ by orders of magnitude, and the caveat must travel with the label: NAS suppl…

★ For the kids: A battery that has to stay red-hot inside to work. It has been quietly running on power grids for over twenty years, and almost nobody talks about it.

Technical: Sudworth, J.L., Journal of Power Sources 100, 149-163 (2001); NGK Insulators NAS deployment disclosures (2024) — vendor-tier, corroborated by public grid-connection records. Binding caveats: ~75-85% round-trip efficiency, parasitic heating, high cell cost, and the 2011 Tsukuba fire that forced module fusing/isolation redesign. ZEBRA is niche (telecom/defence backup). Annual additions are small: viable attaches to the TECHNOLOGY's field record, never to its growth rate.

Frozen prediction: Through 2028-12-31 cumulative installed high-temperature sodium (NAS) energy capacity worldwide remains at least 5x cumulative installed iron-air capacity, AND no second independent NAS producer reaches >=100 MWh/yr.

SPLIT out of the founding lfp-lmfp lump and entered at modeled. LiMn(x)Fe(1-x)PO4 raises the olivine plateau from 3.4 V to ~4.1 V for ~15-20% more energy at similar cost and is the most-cited 'LFP sequel', but its evidence is not LFP's: Mn3+ Jahn-Teller distortion, Mn dissolution into the electrolyte and two-plateau conductivity loss keep cycle life below LFP, and shipped volumes are pilot-to-early-series against LFP's terawatt-hours. It also re-imports a high-purity manganese-sulphate refining dependency the lumped record hid.

★ For the kids: A tweak to the cheap iron battery that adds manganese for more punch — promising, but it wears out faster and is barely being built yet.

Technical: Olivine base: Padhi et al., J. Electrochem. Soc. 144, 1188-1194 (1997); manganese-substitution limits reviewed in Masquelier & Croguennec, Chemical Reviews 113, 6552-6591 (2013). The prior record's only supporting evidence was a 2023 cell announcement — announcement tier. Cross-links high-purity-manganese-sulfate (open): an LMFP win tightens battery-grade HPMSM, where China retains >85% of production.

Frozen prediction: LMFP (>=20% Mn in the olivine) remains below 10% of global LFP-family cathode shipments through 2028-12-31.

HELD open. Lithium carbonate ~+37% YTD roughly doubles LFP black-mass lithium value vs the 2025 trough, but no operator has disclosed a positive gross margin at >5 kt/yr without gate fees or credits.

★ For the kids: Old iron-phosphate batteries have little valuable metal except lithium. Lithium got pricier in 2026, so recycling them might finally pay - nobody has shown the numbers yet.

Technical: At ~$23k/t Li2CO3, contained lithium in LFP cathode powder is worth ~$3,500-4,500/t before losses vs ~$1,500-1,800/t hydromet cost - positive on paper, unproven in any audited P&L.

Frozen prediction: Through 2027-12-31 no operator reports a positive gross margin on LFP-only black-mass hydromet at >5,000 t/yr absent gate fee, subsidy or credit; author P(holds) 0.65.

TIER HELD at modeled. Whether black mass is 'hazardous waste' or 'a metal-bearing intermediate' decides where the world's spent-battery value is refined. The EU classified black mass as hazardous waste effective 2025-03-05, barring export unless sufficient pretreatment and industrial reuse can be demonstrated; Basel COP-17 (Geneva, April-May 2025) did NOT finalise its waste-battery technical guidelines, deferring them to OEWG-15 in 2026 and possible adoption at COP-18 in 2027. The result is a multi-year window of regionally divergent classification in which US black mass, in oversupply, keeps…

★ For the kids: When old batteries are shredded you get a black powder full of metals. Europe just said that powder is dangerous waste; China said the opposite. Where the powder may travel decides where the metals get made into new batteries.

Technical: European Commission Delegated Decision amending the List of Waste under Directive 2008/98/EC (mandatory 2026-11-09); GB/T 45203-2024 plus the MEE/GAC customs announcement. The frequently-quoted ~75% China share of global black-mass processing capacity is DGAP (2025) think-tank tier, not audited, and is excluded from the tier claim. Grading series exists and is public: Fastmarkets NCM cathode black-powder payable CIF China vs CIF Korea/US, plus EU shipment notifications by destination — which is why the prediction is written on DESTINATION SHARE, the one number a press release cannot satisfy.

Frozen prediction: The Basel Convention has not adopted binding technical guidelines specifically governing transboundary movement of lithium-ion black mass before end-2027.

MODELED, duplicate proposals merged. Ion-adsorption clays are the only bulk heavy-rare-earth source class outside hard rock, and Serra Verde's Pela Ema in Goias, Brazil has produced since 2024 as the first Western ionic-clay operation, targeting 6,400 t/yr TREO by end-2027; USA Rare Earth agreed in April 2026 to acquire it for ~$2.8bn (close expected Q3 2026, under Brazilian court challenge). The honest read is that heavy-REE relief arrived through GEOLOGY, not through the exotic separation chemistries this lens has been scoring — but the mine ships MIXED CARBONATE, and separation, the step t…

★ For the kids: Special clay in Brazil holds the rare metals that make magnets work in hot motors. Digging it up is only half the job — the sorting still happens far away.

Technical: Mechanism: Borst et al., Nature Communications 11, 4386 (2020) — heavy REE held as easily leachable outer-sphere hydrated complexes, dominantly on kaolinite. USA Rare Earth Form 8-K / DEFA14A, SEC EDGAR (2026): 15-year offtake covering 100% of Phase 1 Nd/Pr/Dy/Tb with price floors. Concentrate Dy2O3/Tb4O7 grades circulating in trade blogs are not independently sourced and are excluded. Context: USGS MCS 2026 records China's April-2025 controls (Sm, Gd, Tb, Dy, Lu, Sc, Y) and the October-2025 extension (Eu, Ho, Er, Tm, Yb).

Frozen prediction: Through 2027-12-31 at least 80% of Western ionic-clay mixed rare-earth carbonate output is separated outside the US and EU; no ex-China ionic-clay operation sustains >=5,000 t/yr TREO for four consecutive quarters; and no separated Dy or Tb oxide is produced in Brazil.

OPEN. No commercial claystone lithium plant operates anywhere and none ever has. Thacker Pass Phase 1 (40 kt/yr LCE, McDermitt caldera) is under construction with detailed engineering >93-95% complete and first production guided to late 2027 (a slip from earlier expectations) on 2026 capex of $1.3-1.6bn; Rhyolite Ridge is the other first mover. The flowsheet — sulphuric acid leach of clay, neutralisation, impurity removal, carbonation — is first-of-a-kind at scale, and by the FOAK base rate the binding steps are acid consumption, solid-liquid separation of fine clay residue and materials of c…

★ For the kids: Some rocks are like lithium-soaked mud instead of hard stone or salty water. Nobody has ever run a big factory on lithium mud, so we do not know if it works.

Technical: Resource geology: Benson, Coble, Rytuba & Mahood, Nature Communications 8, 270 (2017). Schedule per Lithium Americas 2026 SEC disclosures and the 2026-02-19 project update; $2.23bn DOE loan, GM 38% JV, Bechtel EPC. In clay leaching the residue, not the lithium, governs plant throughput. A rising lithium price (NE Asia ~USD 21/kg June 2026) both funds and forgives a first-of-a-kind flowsheet — exactly when scale-up discipline slips.

Frozen prediction: No claystone/sedimentary lithium operation produces commercial battery-grade lithium chemical before 2027-10-01, produces >=20,000 t LCE in any rolling 12-month period through 2028-12-31, or exceeds 50% of nameplate in its first full production year.

OPEN, for a blunt reason: the end-of-life feedstock does not exist. Roughly 7-8 t/yr of primary iridium supply, essentially all a by-product of South African PGM mining, faces PEM anode loadings a multi-GW buildout cannot cover — but 2026 is expected to be the first year of commercial-scale iridium use in PEM electrolysis at all, so the earliest meaningful EOL stacks are a decade out. Today's 'iridium recycling' is crucible, spark-plug and manufacturing-scrap recovery; refiners (Johnson Matthey's new UK plant, Heraeus/McCol in Canada) are building capacity ahead of a stream that has not arriv…

★ For the kids: Iridium is one of the rarest metals on Earth and hydrogen machines need it. We would love to reuse it — but the machines using it are brand new, so there is almost nothing old to recycle.

Technical: Johnson Matthey PGM Market Report 2026 (iridium and ruthenium in deficit; 2026 flagged as first commercial-scale PEM iridium use) and the JM HiSPEC thrifting roadmap toward ~80 kg/GW by 2030, with a ~30 kg/GW long-term aspiration. The World Platinum Investment Council's contrary position — that iridium is not a hard bottleneck on its projections — is a genuine live disagreement and is itself part of the case for the open tier.

Frozen prediction: No end-of-life PEM-electrolyzer iridium recycling stream exceeding 100 kg/yr is publicly reported anywhere before 2028-12-31.

OPEN, catalogue-gram scale — split out of the generic solid-electrolyte domain so a sulfide scale-up can never carry halide evidence. Halides reach ~1-2 mS/cm with >4 V oxidative stability and pair with high-voltage cathodes WITHOUT the coatings sulfides require and without sulfide's H2S hazard, but they sell in 10 g - 1 kg research quantities, are moisture-sensitive, and reduce at the anode. The supply consequence is branch-specific and counterintuitive: an In/Y/Sc-based halide win would CREATE an indium chokepoint (indium is a zinc by-product) rather than relieve one; Zr versions would not.

★ For the kids: Some battery 'jelly' is made from salt-like chlorides. It handles high voltage well and gets along with the other parts, but it hates humidity — and you can only buy it by the spoonful.

Technical: Asano et al., Advanced Materials 30, 1803075 (2018) established Li3YCl6/Li3YBr6; Li et al., Angew. Chem. Int. Ed. 58, 16427-16432 (2019) reported water-mediated Li3InCl6 at ~2 mS/cm; Bouguern et al., Advanced Materials (2026), doi:10.1002/adma.202513255 compares sulfide vs halide commercialisation. Merchant supply is specialty-catalogue quantity (NEI Corporation listings, 2026): the tier is set by that tonnage, not by the conductivity number. 'Commercially available' from a catalogue is not merchant supply.

Frozen prediction: No cell using a halide (Li3MCl6-type) electrolyte as its primary separator reaches >=100 MWh/yr production before 2029-12-31, enters series production in any road vehicle before 2030-12-31, or drives merchant halide-electrolyte supply above 100 t/yr through 2028-12-31.

OPEN. Li7La3Zr2O12 is the air-stable, lithium-metal-tolerant oxide route, and the only solid-electrolyte family whose scale-up would CONSUME a stranded co-product — lanthanum, the surplus light rare earth co-mined with NdPr — rather than tighten a scarce input. The gates are manufacturing, not chemistry: ~1,100 C sintering with lithium loss, brittle thin-film handling, and dendrite penetration above a critical current density. Only ~5% of announced solid-state programmes use LLZO and no LLZO cell is in series production, so the tier is set by absent tonnage.

★ For the kids: This battery filling is a hard ceramic made partly from lanthanum — a metal miners already dig up and struggle to sell.

Technical: Murugan, Thangadurai & Weppner, Angew. Chem. Int. Ed. 46, 7778-7781 (2007) established fast Li+ conduction in cubic LLZO. 2025-26 reviews (Inorganics 14, 148, 2026; Chemical Engineering Journal LLZO interfaces review, 2025) put production at pilot scale with a ~30-40% cost premium. Cross-links light-rare-earth-substitution from the opposite direction: LLZO is a lanthanum DEMAND SINK, not a substitution risk — the only asymmetry of its kind in the electrolyte race.

Frozen prediction: No LLZO-based cell reaches >100 MWh/yr installed production before 2029-12-31, and lanthanum consumed by solid electrolytes stays below 1% of annual global La2O3 output through 2028-12-31.

MODELED, duplicate proposals merged. Fluorophosphates (Na3V2(PO4)2F3) and pyrophosphate-phosphates (Na4Fe3(PO4)2P2O7) trade energy density for the two things grid storage actually buys — multi-thousand-cycle life and wide-temperature stability — and Chinese sodium ESS is converging here rather than on layered oxides. Held at modeled: 2026 scale-up reviews still name poor electronic conductivity and pure-phase synthesis as unresolved industrialisation gates, and production is pilot-to-small-series only. Branch matters for the chokepoint: NVPF re-imports a VANADIUM dependency shared with vanadi…

★ For the kids: A salt-battery recipe built from iron and phosphate that can be charged thousands of times — great for storing power, not for long-range cars. One version needs a scarce metal called vanadium.

Technical: Kim et al., J. Am. Chem. Soc. 134, 10369-10372 (2012), doi:10.1021/ja3038646 introduced Na4Fe3(PO4)2(P2O7); framework reference Masquelier & Croguennec, Chemical Reviews 113, 6552-6591 (2013). CITATION CORRECTION carried forward: Barpanda et al., Nature Communications 5, 4358 (2014) is the alluaudite SULPHATE Na2Fe2(SO4)3 at 3.8 V, NOT NFPP — a common secondary-source mis-attribution. Scale-up gates per Green Chemistry (RSC, 2026), doi:10.1039/d5gc06522d.

Frozen prediction: By 2028-12-31 polyanionic (NFPP/NVPF) cathodes exceed layered oxides in shipped sodium-ion stationary-storage GWh, AND no polyanionic sodium cell line reaches >=1 GWh/yr operating capacity outside China.

MODELED, and the counterintuitive lithium story worth keeping. Silicon and hard-carbon anodes lose 10-25% of lithium inventory on first cycle, so cells put it back — via sacrificial cathode additives (Li2NiO2, Li2O, Li2C4O4) that decompose on first charge, or via lithium-metal powder. Additives are drop-in compatible with existing lines and are the only fielded route; Li-metal powder remains a dry-room handling problem. The consequence corrects a widespread read: silicon adoption destroys GRAPHITE demand, not lithium demand, because part of the energy gain is handed back as sacrificial lithiu…

★ For the kids: New anodes swallow some lithium the first time and never give it back, so engineers pack extra lithium in as a snack.

Technical: Holtstiege, Barmann, Nolle, Winter & Placke, Batteries 4, 4 (2018) reviews the routes. Li2NiO2 at 8 wt% in NCM811/silicon-graphite 6.13 Ah pouch cells gave 80.9% retention after 700 cycles at 1C (Materials Letters, 2023) with ~295 mAh/g irreversible capacity at ~31% first-cycle efficiency. No disclosed mass-market Li-metal-powder line. Falsifier for the demand read: a disclosed Si-dominant commercial cell whose lithium content per kWh is below a comparable graphite/NMC cell.

Frozen prediction: By 2028-12-31 no cell line above 1 GWh cumulative output publicly discloses metallic-lithium-powder prelithiation; sacrificial cathode additives remain the only fielded route.

TIER HELD at open, and the record's handling of its own promotion gate is correct and should be preserved: the gate reads '>=1 Ah room-temperature cell with independently reported cycle life at practical N/P ratio and lean electrolyte', and the verified Nature 651, 383-389 (2026) result satisfies format (13 Ah), anode thinness (20 um Li) and electrolyte leanness (0.48 g/Ah) in one peer-reviewed device. The two load-bearing clauses fail: cycle life is not clearly reported (the widely repeated ~115-cycle figure is unverified here), and the result is single-group. One correction carried from the…

★ For the kids: Pure lithium metal would store the most energy of all, but at normal temperatures it grows spikes that wreck the battery. Nobody sells one for a car.

Technical: Related tracked domains: anode-free-cells (open), high-entropy-battery-materials (open, anode-free NMC622||Cu pouch work), lithium-metal-polymer-cells (modeled, 60-80 C). Any promotion requires a >=1 Ah room-temperature cell with independently reported cycle life at practical N/P ratio and lean electrolyte, not a coin-cell coulombic-efficiency number.

Frozen prediction: No room-temperature lithium-metal-anode cell ships in a series-production passenger EV before 2029-12-31, no >=1 GWh/yr line is commissioned for one before 2028-12-31, AND no independent replication of a >=1 Ah lean-electrolyte lithium-metal cell exceeding 300 cycles appears before 2028-12-31.

VIABLE, and added deliberately as the honest BENCHMARK for every fast-charge anode claim. Li4Ti5O12 operates at ~1.55 V vs Li, so no lithium plating and no SEI-driven fade: 10,000-20,000 cycles, 6C+ charging and safe sub-zero charge acceptance, at ~40% lower cell energy (~80-90 Wh/kg). It has shipped commercially since 2008 (Toshiba SCiB in EVs, buses, rail and grid) from more than one producer, and its titanium feedstock is unconstrained. Niobium-oxide fast-charge anodes must beat a shipping incumbent, not graphite — a framing that materially lowers the niobium value-capture thesis.

★ For the kids: A battery that charges in minutes and lasts for decades, but holds less energy — and it has been on sale since 2008.

Technical: Ohzuku, Ueda & Yamamoto, J. Electrochem. Soc. 142, 1431-1435 (1995) established the ~0.2% lattice-strain mechanism. The energy penalty is intrinsic (high anode potential), not an engineering gap, which is why LTO never displaced graphite and why its existence is evidence about the CEILING of the fast-charge niche rather than about its growth.

Frozen prediction: Niobium-oxide (NTO/Wadsley-Roth) fast-charge anodes do not exceed lithium-titanate in annual shipped cell capacity before 2028-12-31.

OPEN — entered at the lowest tier, explicitly rejecting the standing 'cobalt-free is commercialised' framing. SVOLT put a zero-cobalt Ni75/Mn25 cathode into series production at Jintan in July 2021 with 5,000 t/yr of material and a launch vehicle, which read as viable at the time. Five years on there is no disclosed sustained volume, and the case rested on a self-claimed ~5% cost saving versus NCM that the LFP price crash erased. Pure LiNiO2's cation mixing and surface reactivity remain unfixed. A one-time launch is not a shipping chemistry.

★ For the kids: A battery recipe that removed the priciest metal entirely, launched once in 2021, and then went quiet.

Technical: Bianchini, Roca-Ayats, Hartmann, Brezesinski & Janek, Angew. Chem. Int. Ed. 58, 10434-10458 (2019), 'There and Back Again — The Journey of LiNiO2 as a Cathode-Active Material', documents the recurring cation-mixing / surface-reactivity failure across five decades — the base-rate document for this domain and a good general prior on chemistries that get rediscovered.

Frozen prediction: No zero-cobalt (Co = 0) layered nickel cathode exceeds 5% of global high-nickel (>=80% Ni) cathode shipments before 2028-12-31.

OPEN, and a clean hype-half-life datapoint. Honda/Caltech/JPL demonstrated the first room-temperature fluoride-ion cell in 2018 after decades of >150 C operation. The publicised '10x lithium energy density' is a theoretical ceiling for metal-fluoride couples, not a measured device, and is retired as a near-term claim. Eight years on there is no multi-Ah format, no independent cycle-life replication, and the enabling electrolyte is itself deeply fluorinated — so the chemistry marketed as an escape from lithium is not an escape from fluorine, and inherits the same EU regulatory exposure as LiFS…

★ For the kids: A battery that moves fluorine instead of lithium. It worked once on a lab bench in 2018 and has not grown up since.

Technical: Davis, V.K., Bates, C.M., Omichi, K. et al., Science 362, 1144-1148 (2018): reversible room-temperature cycling with dry tetramethylammonium fluoride in a fluorinated ether and a Cu@LaF3 core-shell cathode. Note the cross-domain collision worth watching: LaF3 is simultaneously this cathode material and a standard UV/IR optical coating with a small specialised supply base.

Frozen prediction: No fluoride-ion cell of >=1 Ah format with >100 reported cycles appears in the peer-reviewed literature before 2029-12-31. Flagged NON-SCORING under the dormant-technology rule: the field has been static since 2018.

OPEN, and mis-sold. 21 mol/kg LiTFSI widened the aqueous stability window from ~1.23 V to ~3 V, enabling a genuinely non-flammable lithium-ion chemistry — the origin of the 'safe water battery' narrative. But the enabling ingredient is an extreme molality of a fluorinated imide salt: per kWh these cells carry MORE fluorinated salt than the LiPF6/carbonate systems they claim to displace. Non-flammable is not fluorine-free. Salt cost plus anodic aluminium-current-collector corrosion is why no commercial cell exists eleven years on.

★ For the kids: A battery that runs on water — except the water is so packed with special salt that the salt, not the water, is the expensive part.

Technical: Suo, L., Borodin, O., Gao, T. et al., Science 350, 938-943 (2015): ~3 V window at 21 mol/kg LiTFSI with a LiF-rich interphase as the operative mechanism. LiTFSI is PFAS-class, so this domain's EU regulatory exposure matches LiFSI's and its upstream exposure is the fluorspar/HF chain, not lithium.

Frozen prediction: No water-in-salt (>15 mol/kg fluorinated-salt) aqueous cell ships as a commercial product before 2029-12-31. Flagged NON-SCORING: static since 2015.

OPEN, opened deliberately at the lowest tier as a pre-emptive guard. The Kneller-Hawig concept (1991) — coupling a high-anisotropy hard phase to a high-magnetization soft phase to beat the (BH)max of either — has been 'about to double magnet energy product' for 35 years and has never produced a bulk anisotropic commercial magnet. Honest state of the art: ~14 MGOe in ISOTROPIC nanocomposites at room temperature; higher values (~310 kJ/m3) appear only in optimized/idealized combinations, never in manufacturable bulk. Entered at open specifically so no future press release can enter this study a…

★ For the kids: Imagine gluing a very stubborn magnet to a strong-but-floppy one so the stubborn one holds it in line. It works beautifully at a few nanometres — then you squash millions together, the pieces grow, and the trick stops working.

Technical: Exchange coupling requires the soft-phase dimension below roughly twice the hard-phase domain-wall width (~5-10 nm). Producing that microstructure at powder scale is achievable; preserving it through texturing and hot compaction is not — grain growth during consolidation destroys the length scale the effect depends on (Magnetochemistry 11(11), 99, 2025; Heliyon technology analysis, 2024). Promotion bar: a bulk ANISOTROPIC exchange-spring magnet (Br/Js > 0.85) with a published full demagnetization curve at >=100 g scale.

Frozen prediction: No bulk anisotropic exchange-spring magnet with Br/Js > 0.85 and a published demagnetization curve at >=100 g scale is reported in peer-reviewed literature by 2028-12-31.

MODELED. The largest un-worked lever in magnet materials efficiency is yield, not chemistry: roughly 20-30 wt% of sintered NdFeB is lost as swarf during cutting and grinding, and that swarf carries ~30 wt% rare earth — richer than any mined ore. Two routes attack it: near-net-shape/additive forming of BONDED magnets (injection-moulded bonded magnets are commercially shipping; ORNL's BAAM and binder-jetting routes are pilot) and short-loop swarf capture (VAC contracting 100% of its Sumter by-product to a dedicated recycler under a 10-year agreement). The AM route and the closed loop are pilot-…

★ For the kids: When you carve a magnet into shape, up to a third ends up as dust on the floor — and that dust is richer in rare earths than any mine. Either print the shape, or sweep up the dust.

Technical: Li, L. et al. (ORNL), Scientific Reports 6, 36212 (2016): 65 vol% isotropic NdFeB in PA-12, properties comparable to injection-moulded, no tooling, near-zero machining waste. AM bonded magnets are isotropic and polymer-diluted at ~7-10 MGOe against ~50 MGOe for sintered anisotropic NdFeB — they do NOT substitute for traction rotor magnets and must never be scored as if they do; their value is tooling elimination and graded/Halbach geometries that cut total magnet mass at constant airgap flux. Swarf is oxidized and pyrophoric so it routes hydrometallurgically (~97% REE recovery at >99.5% oxide…

Frozen prediction: By 2027-12-31 no automotive traction-motor rotor magnet in series production is made by additive manufacturing; AM magnets remain confined to sensors, actuators and prototypes.

OPEN, emphatically. Replacing permanent-magnet excitation with a REBCO field winding removes NdFeB from the rotor entirely, which is relevant to the two applications driving NdFeB demand growth (offshore-wind direct drive and prospective aviation propulsion). The anchor evidence is a single full-scale field test: the EU EcoSwing 3.6 MW REBCO generator on a real turbine, 40 rotor coils consuming >20 km of coated conductor, TRL 7 — one demonstrator, seven years ago, with no commercial follow-through.

★ For the kids: Instead of a permanent magnet you can use a coil of wire with zero resistance when it is very, very cold. One real wind turbine ran on one in 2019 — but you must keep it cold forever, and one generator eats twenty kilometres of special wire.

Technical: Bergen, A. et al. (EcoSwing), Supercond. Sci. Technol. 32 (2019). REBCO is 'rare-earth-free' only in the magnet sense — the tape is a rare-earth barium copper oxide (Y or Gd) on Hastelloy with silver and copper stabilizer, so it swaps an NdPr/Dy chokepoint for a coated-conductor and cryogenics chokepoint. Engineering blockers: rotating cryostat torque tubes and current leads, quench protection in a rotating frame, and AC loss under transient/fault conditions.

Frozen prediction: No non-demonstration REBCO rotating machine enters series production in a vehicle, ship, aircraft or industrial drive before 2028-12-31.

MODELED. REBCO coated conductor is a genuine commercial product from roughly 15 makers at ~5,000-10,000 km/yr of 12 mm-equivalent global capacity, and it enabled the ~20 T SPARC toroidal-field model coil — but the applications it is priced on need an order of magnitude more tape than the world makes, since one prototype fusion device can consume ~10,000 km of 4 mm tape. Kept at modeled rather than promoted on 'commercially shipping': the tier here is being asked to carry a supply claim, and the supply is the thing that fails.

★ For the kids: These thin metal ribbons carry electricity with no loss and make super-strong magnets. The problem is the world makes only a few thousand kilometres a year, and one machine can eat all of it.

Technical: Hartwig, Z.S. et al., IEEE Trans. Appl. Supercond. 34 (2024) for the magnet demonstration; 'Commercial compact fusion triggered REBCO tape industry: pulsed laser deposition technology opportunities and challenges', Superconductivity (Elsevier, 2025) for capacity — four PLD makers (Faraday Factory Japan, Shanghai Superconductor, Fujikura, S-Innovation) supply over half of world output. Every tape carries a silver overlayer over a rare-earth barium copper oxide film, so it adds to the chokepoints it is supposed to bypass.

Frozen prediction: Through 2028-12-31 world REBCO tape capacity stays below 50,000 km/yr (12 mm-equivalent).

MODELED, duplicate proposals merged. With LME copper near record levels (>$13,000/t, January 2026), aluminium is the elastic answer on the conductor side: it is already the default in overhead transmission and increasingly standard in distribution cable, busbars and cast induction rotors. It is NOT proven in traction-motor stator windings — aluminium hairpin designs remain prototype (the UK Alumotor consortium's rare-earth-free synchronous reluctance motor, claiming up to 60% manufacturing cost reduction). Physics caps the wedge-relevant use, so the tier is set by the prototype, not by the tr…

★ For the kids: Copper is the best wire we have but costs a lot. Aluminium is cheaper and lighter, but electricity squeezes through it less easily, so you need a fatter wire. It works in long power lines; car motors are still copper.

Technical: Aluminium is ~61% IACS at ~30% of copper's density, so equal DC resistance needs ~1.64x cross-section: it wins conductance per kilogram and loses conductance per volume, making substitution space- and joint-limited rather than physics-limited. Costs ~1.5-2% motor efficiency at equal geometry, with creep and galvanic corrosion at Cu-Al joints as the historical failure mode. Nottingham-group optimizations (IEEE, 2023-2025) find aluminium hairpins Pareto-competitive only in utility/low-cost vehicle classes. The coupling that makes this lens-relevant: removing rare-earth magnets (EESM/SynRM) INCR…

Frozen prediction: Through 2028-06-30 no top-10 global automaker starts series production of a BEV traction motor with aluminium stator windings; aluminium gains stay confined to transmission, distribution cable, busbar and harness.

MODELED. PV metallisation consumed roughly 8,600 t of silver in the 703 GW shipped in 2024, and TOPCon still uses ~10-12 mg/W. Per-watt thrifting is real and shipping (silver-coated copper paste on rear contacts at LONGi and Jinko; ITRPV 2026 projects TOPCon to ~6.3 mg/W and HJT to ~4.3 mg/W over ten years), but full copper electroplating reaches ~1.1 mg/W only in the lab and needs new plating tools, not a screen-printer swap. The skeptic point that decides the investment read: falling per-watt intensity does not imply falling total silver demand while shipments grow — the volume term has his…

★ For the kids: Solar panels use a lot of silver to carry electricity. Engineers are learning to use less, and to try copper instead — but copper printing is not in big factories yet.

Technical: VDMA ITRPV 2026 edition for the intensity roadmap; Fraunhofer ISE reported a ~10x silver reduction via plated Ni/Cu/Ag contacts (2025-26); LONGi stated it would begin substituting base metals in Q2 2026. Plating durability under deep thermal cycling is the unproven part and is the correct pass/fail gate, not the mg/W number.

Frozen prediction: Through 2028-12-31 copper-plated metallisation stays below 5% of global c-Si module shipments, and annual PV silver demand does not fall more than 20% below its peak year despite falling per-watt intensity.

VIABLE. Silicon-carbide MOSFETs ship commercially in EV traction inverters and 800 V charging (since the 2018 Model 3 inverter) from multiple vendors with 200 mm substrate lines running, and GaN power ICs ship in consumer chargers and are entering 48 V rails — commercial shipping, multi-producer, replication-grade. This is a materials advance acting on the battery chokepoint from the DEMAND side: several percent lower inverter loss is fewer kWh per mile. The supply read is counterintuitive and worth freezing: power GaN uses gram-scale gallium, so gallium export controls bind LED/RF/substrate …

★ For the kids: New chips made from tougher materials waste less electricity, so an electric car goes further on the same battery.

Technical: Millan, Godignon, Perpina, Perez-Tomas & Rebollo, IEEE Transactions on Power Electronics 29, 2155-2163 (2014); commercial status per 200 mm SiC substrate ramps (Wolfspeed, STMicroelectronics, Infineon, 2022-2025). Sintered SiC is also the space-telescope mirror substrate material (ESA Herschel, 3.5 m) — the same industrial base, which makes the telescope bridge a shared supply chain rather than an analogy.

Frozen prediction: Through 2028-12-31 no documented allocation shortage of GaN POWER devices is attributed to gallium export controls; any documented attribution falsifies the low-gallium-intensity read.

OPEN — the fuel-cell twin of the iridium-in-electrolyzers problem. PEM fuel cells still need platinum on the cathode; iron-nitrogen-carbon catalysts remove PGMs entirely and reach useful initial oxygen-reduction activity, but durability collapses within hundreds of hours under automotive load cycling, and no commercial stack ships a PGM-free primary cathode catalyst fifteen years after the founding result. The commercially real lever has been platinum THRIFTING (loadings cut roughly an order of magnitude since the 1990s), not replacement — the same dematerialization-beats-substitution pattern…

★ For the kids: Hydrogen cars need a little platinum; scientists are trying to swap it for iron, but the iron version wears out too fast.

Technical: Wu, G., More, K.L., Johnston, C.M. & Zelenay, P., Science 332, 443-447 (2011) established the Fe/Co-N-C class; every subsequent review names durability, not activity, as the blocker. The generalisable prior: across three separate PGM/rare-earth domains in this lens, thrifting has delivered every realised reduction and substitution has delivered none.

Frozen prediction: No commercially sold PEM fuel-cell stack ships with a PGM-free (Fe-N-C class) cathode catalyst as its primary ORR catalyst before 2029-12-31.

TIER HELD at viable, but scope-bound and two supporting claims struck. Merged with the duplicate proposal `battery-separators-coated`. Viable rests ONLY on the separator technology: wet-process polyolefin microporous film with alumina/boehmite coating has shipped at terawatt-hour scale for over a decade from at least six mutually independent producers (Asahi Kasei/Celgard, SEMCORP, SK ie technology, Toray, W-Scope, ENTEK). That is commercial shipping and replication-grade. STRUCK FROM TIER SUPPORT: (a) the 'the United States has no wet-process base-film plant at all' chokepoint framing — ex-C…

★ For the kids: Every battery has a thin plastic sheet keeping its two halves apart. Making it well is a mature, proven business — but almost every company that knows how is in Japan, Korea or China, and only one is American-owned.

Technical: Standing references: Arora & Zhang, Chem. Rev. 104, 4419-4462 (2004); Lee, Yanilmaz, Toprakci, Fu & Zhang, Energy Environ. Sci. 7, 3857-3886 (2014). US DOE LPO conditional commitment of up to $1.2B to ENTEK for Terre Haute (2024) plus the October 2022 $200M grant; ENTEK is the only US-OWNED wet-process producer, not the only wet-process plant reachable by US buyers. Market-research share tables (Mordor, Fortune, GlobeNewswire syndications) and the SMM reference-price series are excluded from the tier claim. Inherits delivery risk from battery-materials-export-control-thresholds (MOFCOM No. 61…

Frozen prediction: ENTEK Terre Haute does not reach >=50% of its stated ~1.4 bn m2/yr nameplate for two consecutive quarters before 2028-12-31, AND no second US-owned wet-process separator producer begins commercial revenue production before then. The SMM price clause of the original prediction is struck as ungradeable (see retired: smm-price-series-as-frozen-metric).

DEMOTED viable -> modeled. Every quantitative leg of the viable claim traces to one commercial trade-data provider (SMM, 2025-01-22): the producer list (Fulin Precision, Hunan Yuneng, Lopal, Defang Nano, Brunp), the ~3,000 yuan/t premium, the ~800,000 t 2025 high-compaction demand, the 4.48 Mt total nameplate, and the CATL prepayment. Five producer names inside one unverifiable source is one source, not five — this is the same error the ledger's own single-producer-replication bar exists to catch, committed at the level of the SOURCE rather than the producer. The one genuinely independent cor…

★ For the kids: The cheapest battery recipe did not change; makers learned to squeeze the powder harder. That the squeezed-tight kind sells for more is something one market-data company says, and nobody else has checked.

Technical: Base chemistry: Padhi, Nanjundaswamy & Goodenough, J. Electrochem. Soc. 144, 1188-1194 (1997). Generation bands (3rd gen 2.4-2.5, 4th gen >=2.6, 5th gen pilot) are SMM-attributed. Independent leg: MOFCOM/MOST catalogue amendment 2025-07-15, LFP/LMFP cathode process technology restricted at >=2.58 g/cm3. Promotion gate set before the fact: a second independent source — an audited filing, a named customer disclosure, or a second price-reporting agency — corroborating both the grade and the premium.

Frozen prediction: No plant outside China publicly confirms commercial production of >=2.6 g/cm3 compaction-density LFP cathode powder at >=10,000 t/yr before 2027-12-31, AND no second independent source corroborates a >=2,000 yuan/t high-compaction premium before 2027-12-31.

HELD at modeled with the dated ladder: MOFCOM/GACC No. 23 (2023, Ga/Ge), No. 39 (2023, graphite), No. 46 (2024, Sb/superhard), 3 Dec 2024 US-specific Ga/Ge/Sb ban, No. 18 (Apr 2025, seven medium/heavy REEs + magnets), Oct 2025 Nos. 61/62 (RE processing tech, extraterritorial), Nov 2025 one-year suspension of the October measures. Net effect: licence-latency, stockpiling and price term-structure, not physical shortage. USGS MCS 2025: no US primary gallium; ex-China Dy/Tb prices diverged from domestic. ENGINE leads 45e2721435d1, 82a74515d023, de5c5f5cc50e read only as questions. Federal Registe…

★ For the kids: China made it harder to sell some special metals and magnets abroad, then paused part of that. Prices outside China jumped, but new mines and factories elsewhere are still slow.

Technical: Licensing covers NdFeB grades containing controlled HREEs, so the practical hit is on H/SH/UH traction grades; substitution pressure falls first on GBD thrifting and HRE-free hot-deformed magnets.

Frozen prediction: Ex-China dysprosium oxide quote remains >=30% above the Chinese domestic quote on 2026-12-31 (p~0.6).

TIER HELD at open; no correction needed and the record is a model of how to enter a domain. Single-group lab work (Jaschin, Tang & Wachsman, ACS Energy Lett. 10, 2610-2616 (2025)): Zn,Mg dual-doped NASICON separators at 18-37 um in a porous-dense bilayer, sodium metal infiltrated, Na||Na3V2(PO4)3 cycled at 22 C to 10.8 mA/cm2 (6C) at 1.8 mAh/cm2. The record correctly refuses to let beta''-alumina's 300 C NAS/ZEBRA shipping history or the sulfide-led lithium solid-state pilot lines lend evidence, and correctly records the widely-repeated 286 Wh/kg as an architectural projection rather than a m…

Frozen prediction: No cell using a room-temperature sodium solid electrolyte reaches >=100 MWh/yr production before 2029-12-31, AND no group independent of the originating laboratory reports >=10 mA/cm2 room-temperature cycling in a >=1 Ah format before 2028-12-31.

TIER HELD at open; MERGED from three duplicate proposals (`composite-and-ultrathin-current-collectors`, `composite-metallized-current-collectors`, `composite-polymer-current-collectors`) that were one object described three times. The merged record keeps the single most useful distinction any of the three drew: Route A, polymer-cored composite foil (~1 um Cu or sub-micron Al on a 4-6 um PET/PP/PI core), which cuts collector metal mass by roughly two-thirds and adds a genuine fuse — the polymer melts and severs the current path at an internal short; and Route B, simple thinning of electrolytic…

★ For the kids: Batteries use a thin copper sheet inside. One idea replaces most of the copper with plastic; another just makes the copper thinner. The plastic idea gets the headlines every year since 2022, but the thinner copper is what factories actually use.

Technical: Reference geometry from the literature: ~750 nm nanocrystalline Al on 6 um PET replacing 14 um Al foil; ~1 um Cu per face on 4-6 um PET/PP. Distinct third object NOT to be merged in: 3D porous copper foil for solid-state anodes, which has no mass production anywhere and belongs with the lithium-metal records. Chinese supplier capex announcements (e.g. Nord's ~RMB 2.5bn programme) are company-stated and excluded. Promotion gate: a disclosed CELL-MAKER qualification at >=1 GWh/yr, not a foil-maker capacity announcement.

Frozen prediction: Through 2028-12-31 no cell maker discloses >=1 GWh of cells shipped with polymer-cored composite current collectors in an audited or regulator-filed document, AND no named independent teardown organisation confirms composite-foil collectors in a mass-production passenger-EV pack; ultra-thin (<=4.5 um) all-metal copper foil remains the dominant in-cell copper-intensity reduction route through 2027-12-31.

TIER HELD at modeled. This is the best structural catch in the submission: Western anode strategy is denominated in graphitisation furnace capacity — the visible, capital-intensive, subsidy-eligible step — while the binding constraint one level up is precursor. Battery-grade synthetic graphite needs low-sulfur needle coke or qualified petroleum/coal pitch, supply of which is overwhelmingly Chinese and Japanese, contested with the electric-arc-furnace graphite-electrode market, and slow to qualify because coke crystallinity and sulfur set anode first-cycle coulombic efficiency. Modeled rather …

★ For the kids: Battery anodes are made by baking a tar-like coke at enormous heat. Everyone counts the ovens, because ovens are big and photograph well. Almost nobody counts where the coke comes from — and most of it comes from the country the ovens were meant to replace.

Technical: Market structure: Oxford Institute for Energy Studies, Insight 164, 'Western Battery Graphite Markets' (February 2025). Mechanism: needle coke's highly anisotropic, low-sulfur mesophase governs graphitisability at 2,800-3,000 C; substituting isotropic petroleum coke degrades reversible capacity and first-cycle efficiency, so 'we will just use another coke' is a qualification problem, not a procurement one.

Frozen prediction: Through 2028-12-31 China plus Japan retain >=80% of global needle-coke production capacity, AND no North American or European greenfield needle-coke facility dedicated to battery anode precursor reports >=50,000 t of actual annual output in any calendar year.

TIER HELD at modeled. The cleanest live instance of chokepoint migration in the file, and it now has an organisation-tier number: the IEA's Global Critical Minerals Outlook 2025 puts China at ~75% of world purified phosphoric acid and ~98% of LFP cathode material, and explicitly names PPA and high-purity manganese sulphate as the two EMERGING chokepoints where project pipelines point to major supply gaps. The shape is identical to manganese: the ORE is abundant and widely held (USGS puts world capacity on a path from 65.0 Mt P2O5 in 2024 to 70.6 Mt by 2028) while the binding step is purificat…

Frozen prediction: Through 2028-12-31 China retains >=65% of world purified-phosphoric-acid production, AND no North American or European plant supplies >=50 kt/yr of battery-grade PPA into a qualified LFP cathode line.

TIER HELD at modeled, and the record deserves credit for entering at modeled rather than viable in the first place. Multi-wall CNT paste has largely displaced carbon black as the conductive additive in Chinese cells; the wedge-relevant part is SINGLE-wall CNT, which at 0.1-0.5 wt% forms a percolating elastic network that survives silicon's volume swing. Published cell data are stark: a 92% silicon anode retained roughly 150 mAh/g after 100 cycles with no CNT versus roughly 1000 mAh/g at 0.25 wt% and ~1300 mAh/g at 0.5 wt%. Modeled because industrial SWCNT supply traces to essentially one orig…

Frozen prediction: Through 2027-12-31 no producer outside the OCSiAl licence chain publicly discloses >=500 t/yr of SINGLE-wall CNT battery-additive production actually in operation; multi-wall CNT volumes do not count.

TIER HELD at open. Graphite is roughly half the mass of a lithium-ion anode and the most import-concentrated battery material there is, yet virtually every commercial recycler treats it as a combustible carrier and burns or discards it while chasing nickel, cobalt and lithium. Laboratory work shows regenerated black-mass graphite at ~387 mAh/g at 0.1C in half cells — on par with commercial battery-grade — via acid leaching for Al/Cu removal plus mild pyrolysis to strip PVDF, with froth-flotation routes reporting >96% yield at >99.6% purity. This is bench work on industrially sourced feed, not…

Frozen prediction: Recycled anode-grade graphite remains below 1% of global anode active-material supply through end-2028.

TIER HELD at open, and the record's use of dating as evidence is exactly right: the founding Nature Energy demonstration is from 2019, and seven years on there is still no commercial DES leach circuit anywhere. The scale-up blockers are identified and unmoved — DES viscosity orders of magnitude above aqueous acid throttling mass transfer, thermal and oxidative degradation of the hydrogen-bond donor across recycles, and the absence of a cheap selective stripping step to regenerate the solvent. These are the same problems that kept ionic-liquid hydrometallurgy out of industry, which is the corr…

Frozen prediction: No commercial battery recycler operates a deep-eutectic-solvent leach circuit processing 1,000 t/yr or more of black mass before end-2029.

HELD open. Nature Reviews Materials 2023 and Energy Material Advances 2025 conclude low intrinsic metal value makes SIB hydromet recycling unprofitable without regulation; Nature Energy 10, 404-416 (2025) frames SIB competitiveness without a recycling credit. No commercial SIB recycling line reported anywhere.

Frozen prediction: By 2027-12-31 no recycler reports a dedicated sodium-ion black-mass line >1,000 t/yr (P=0.85).

HELD. Chemistry settled and commercially shipping, but essentially all from Asia. UBE Waggaman LA (100 kt/y DMC + 40 kt/y EMC) broke ground Feb 2025, operations targeted Nov 2026; Tinci Baytown broke ground Nov 2025. Zero Western merchant DMC/EMC tonnes shipped as of 2026-07, so the ex-China branch is announced-and-under-construction.

★ For the kids: Batteries are soaked in a special liquid. Almost all of it is made in one part of the world; other countries are only now building their first factories.

Technical: Solvation chemistry per Xu, Chem. Rev. 104, 4303 (2004). Tier scores the ex-China chain, not the chemistry.

Frozen prediction: UBE Waggaman does not report sustained output at or above 100,000 t/y DMC nameplate before 2028-06-30.

HELD, and MERGED with the duplicate silane-precursor-silicon-anodes record. The silane-scarcity thesis is falsified by the supplier's own filings: REC Silicon sold 2,194 MT in 2025 against 7,400 MT Butte nameplate, down from 2,561 MT in 2024, with an impairment and a workforce reduction. Binding constraints are CVD reactor throughput, scaffold carbon supply and OEM qualification. The narrow true claim: US anode-grade silane co-located with the anode plants was removed (Moses Lake shut) and its replacement (Group14 BAM-2, 7,200 t/y, up to $200M DOE award) is under construction with production …

★ For the kids: You spray a silicon gas into a carbon sponge to make a better battery part. People thought the gas would run out; there is plenty of gas, and the hard part is the ovens and the sponges.

Technical: REC Silicon ASA Q4 2025 report (Butte revenue USD 78.1M vs 140.7M; management: silicon-anode transition 'remained delayed'). Deposition scale-up per Xiao et al., Energy Environ. Sci. 18, 4037 (2025). Single-pass silane utilisation is undisclosed by every operator, so silane-per-tonne-of-anode cannot be computed from public data and is NOT asserted.

Frozen prediction: REC Silicon silicon-gas sales stay below 3,500 MT in each of FY2026 and FY2027; and Group14 BAM-2 does not disclose (nor any third party verify) >=3,600 t silane over any twelve-month period ending before 2027-12-31.

HELD, scope stated. The chemistry is commercially shipping at enormous scale in China; the tier scores the EX-CHINA branch, which is demonstration-scale (Austvolt) and pre-feasibility (Sandouville/GalliCam). Reactor know-how, not ore, is the scarce input, and the IEA names cathode precursors as particularly disruption-exposed.

★ For the kids: Before you bake the cathode you grow tiny metal snowballs of exactly the right size. Almost everybody who can do it well is in one country.

Technical: Berk, Beierling, Metzger & Gasteiger, J. Electrochem. Soc. 170, 110513 (2023). IEA Global EV Outlook 2026.

Frozen prediction: Non-Chinese pCAM remains below 10% of global nickel-rich pCAM output through end-2028.

HELD at open, with the market-research numbers STRUCK. Lithium-metal and most solid-state designs need wide, pinhole-free foil below ~20 um handled in a dry room, and no commercial-scale supply chain exists. Every circulating supply figure (China nameplate utilisation, a '$1.8bn 2025 market', supplier-qualification claims) traces to SEO-grade market-research vendors with no primary attribution and is inadmissible here.

★ For the kids: A future lithium battery needs a sheet of pure lithium thinner than kitchen foil with no holes. Nobody makes that in big rolls yet.

Technical: Frith, Lacey & Ulissi, Nat. Commun. 14, 420 (2023) and Albertus et al., Nature Energy 3, 16 (2018) both name thin-foil manufacture as an industrial gate rather than an electrochemistry gate.

Frozen prediction: No supplier publicly confirms shipments of >=100 t/yr battery-grade lithium foil at <=20 um before end-2028.

HELD. Coin-cell scale with thick excess sodium and high electrolyte-to-capacity ratios; dendrites, unstable SEI and gas evolution dominate. The 2026 literature still prescribes combined electrolyte + interface + architecture engineering rather than a single working mechanism.

★ For the kids: Instead of storing sodium in a carbon sponge you could use a sheet of sodium metal, but it grows sharp spikes that short the battery.

Technical: Lee, Paek, Mitlin & Lee, Chem. Rev. 119, 5416 (2019); Hou et al., Adv. Energy Mater. 16, e70866 (2026). Distinct from the sodium-solid-electrolytes record, which concerns the separator.

Frozen prediction: No commercially sold room-temperature sodium-metal-anode cell exists before end-2029.

HELD. The strongest result is a single-laboratory coin-cell claim; ambient air adds CO2 and H2O poisoning that pure-O2 testing does not capture. Stays at the bottom tier until independently reproduced at pouch scale.

★ For the kids: This battery breathes air, which makes it light. Real air also has water and carbon dioxide, and those gunk it up.

Technical: Kondori et al., Science 379, 499 (2023). No independent replication above 1 Ah in ambient air.

Frozen prediction: No independent laboratory reports a Li-air pouch cell above 1 Ah delivering more than 100 cycles in ambient air before end-2029.

HELD. Sodium-ion's entire thesis is cost and the anode precursor is where that cost sits. Resin gives the best capacity and initial coulombic efficiency but is expensive and uses toxic monomers; pitch and coal are cheap and lower-capacity; biomass needs pretreatment and carries feedstock variability. Chinese pricing runs ~RMB 50,000-100,000/t mainstream and 150,000-200,000/t high-end. Low ICE forces cathode oversizing or presodiation, eating the BOM advantage before the comparison is made.

★ For the kids: Salt is cheap; the special charcoal that stores it is not. So a sodium battery is only cheaper when lithium is expensive.

Technical: He, Li, Wu & Chou, Adv. Mater. 37 (2025), DOI 10.1002/adma.202506066; Huang & Wang, Materials 19, 2132 (2026). Peer-reviewed consensus names the absence of a low-cost mass-producible precursor as the industrial bottleneck.

Frozen prediction: Through calendar 2028 no published third-party cell price index or teardown shows a full-year average sodium-ion cell price below LFP on a $/kWh basis.

OPENED AT THE LOWEST TIER and bounded. Every published sample is a melt-spun ribbon or ~30-40 nm nanocrystallite characterised by anisotropy field via singular-point detection, not by a demagnetisation curve on a consolidated magnet. Zr raises Ha 7.5 -> 9.5 kOe; Si LOWERS Ha from ~10.7 kOe while raising coercivity through grain refinement — evidence the observed coercivity is microstructural, not intrinsic. Two overlapping-author papers, no mass, no (BH)max, no replication.

★ For the kids: A new magnet recipe of iron, cobalt and boron. So far only tiny flakes, not a magnet you could put in a motor.

Technical: Joshi et al., AIP Advances 16, 025105 (2026); Abbas et al., AIP Advances 16, 025304 (2026). Compare Nd2Fe14B Ha ~75 kOe — an order of magnitude above the family on the quantity that sets the coercivity ceiling.

Frozen prediction: No (Fe,Co)2B-family magnet with room-temperature intrinsic coercivity >=3 kOe AND (BH)max >=10 MGOe on a consolidated bulk sample >=1 g, with a published full demagnetisation curve, appears in peer-reviewed literature by 2028-12-31.

VIABLE as a shipping magnet family, tier bounded to that scope. Produced commercially since the 1930s, fully rare-earth-free, uniquely thermally stable (Curie ~850 C). Disqualified from substitution by coercivity: best directionally solidified high-Ti alnico reaches Hci 2.0 kOe / (BH)max 10.9 MGOe against the ~3 kOe rotor floor. A floor, not a bridge — and it trades a rare-earth chokepoint for a cobalt one at 24-39 wt% Co.

★ For the kids: An old magnet of aluminium, nickel, cobalt and iron. Fine in some machines, too easy to knock out of alignment for a car motor.

Technical: Zhou, Kramer et al., J. Magn. Magn. Mater. (2016). The additive route (Bishop et al., AIP Advances 16, 025038, 2026) is a processing demonstration with no competitive magnet — and needs an N52 NdFeB fixture to work.

Frozen prediction: No alnico-family magnet reaches Hci >=3 kOe AND (BH)max >=15 MGOe in peer-reviewed literature by 2028-12-31, and no series-production passenger-EV traction motor above 10,000 units/yr uses alnico rotor magnets before 2030-01-01.

HELD, scope bounded on purpose. SMCs are commercial in small, auxiliary and high-frequency machines; the TRACKED claim is SMC as traction-motor relief, which is prototype-only and blocked by physics: Somaloy-class SMC saturates near 1.2 T at relative permeability ~500-850 against ~2.0 T and several thousand for non-oriented electrical steel. Lower flux density means more magnet or more copper for the same torque, so the substitution can eat its own gain.

★ For the kids: Pressed iron powder can be shaped into any motor part, but it cannot carry as much magnetism as flat steel sheets.

Technical: Materials 18, 4405 (2025); Mater. Res. Express (IOP, 2024) DOI 10.1088/2053-1591/ad8413; DOE VTO FY2022 report (OSTI 1901991). Complementary to electrical-steel-noes and amorphous-nanocrystalline-soft-magnets, not a duplicate.

Frozen prediction: Through 2028-12-31 no passenger-EV traction motor above 10,000 units/yr uses an SMC stator core as its primary magnetic circuit, and no peer-reviewed SMC grade reports saturation polarisation >=1.6 T with relative permeability >=2000.

MERGED and held at OPEN — the lower of the two proposed tiers, because the claim that this is THE binding ex-China constraint is inference, not measurement (Lynas and MP both cleared it). Monazite runs ~3-14% ThO2 against a 0.05 wt% statutory source-material threshold, cracking sends ~96-98% of contained thorium to residue, and roughly 2.6 kg of thorium arises per tonne of REE. No plant outside China has demonstrated commercial recovery or beneficial use; the demonstrated state of the art is compliant disposal, and recovery chemistry is bench and pilot only.

★ For the kids: The rocks holding the best magnet metals are slightly radioactive. Getting the metals out leaves a radioactive pile nobody outside China has found a use for.

Technical: 10 CFR 40.13(a) (0.05 wt%) and 40.13(c)(1)(vi) (0.25 wt% in RE products) attach to PRODUCTS, not to ore or residues — which is why the residue stream binds. NRC ML22144A419; Appl. Sci. 15, 11403 (2025). Legal thresholds are facts and confer no technology-readiness tier.

Frozen prediction: By 2028-12-31 no rare-earth separation facility outside China discloses commercial-scale recovery or beneficial use of thorium from its own monazite/NORM residue, and no new commercial-scale US facility licensed to accept separated thorium begins operation.

HELD, tracked as a counter-signal. Two genuine substitutions (solid refrigerant for HFC; La/Ce for Gd) inside a device whose cost is dominated by its permanent-magnet assembly — roughly 0.5-10 kg of neodymium per cooling device. Every deployment is a prototype or demonstration cooler. Net effect on the NdPr chokepoint if it ever scales is ADDITIVE.

★ For the kids: You can make a fridge cold with a magnet instead of a gas, but it needs a big expensive magnet — so a greener fridge would need MORE scarce metal, not less.

Technical: Vinod, Arvindha Babu & Wuppulluri, ACS Omega 9, 11110 (2024); Bjork, Bahl, Smith & Pryds, Int. J. Refrig. 34, 1805 (2011). Exists to keep substitution accounting NET of new magnet demand.

Frozen prediction: Through 2028-12-31 no magnetocaloric refrigeration product ships above 100,000 units/yr worldwide, and magnetocaloric cooling consumes less than 0.1% of annual global sintered NdFeB output.

HELD at open because the tier rests on forecasts, not shipped tonnage. Mechanism is specific: ~30-40 sintered NdFeB motors per current humanoid, roadmapped toward ~78 joints, in exactly the compact high-torque thermally-stressed class where ferrite, SmFeN and externally-excited alternatives fail. Near-term tonnage is trivial; the headline numbers are extrapolations to 2040 unit volumes that do not exist. The ENGINE-origin pantry lead on an Optimus magnet-licensing disclosure is explicitly NOT corroboration and is not counted.

★ For the kids: Every robot joint needs a motor and every motor needs a magnet. If robots are ever built in millions they will want the same scarce magnets cars want.

Technical: Adamas Intelligence (2025-2026) unit-count NdFeB analyses, cited as analyst forecast and raising no maturity. The falsifier is written against measurable near-term share, not the 2040 story.

Frozen prediction: Through 2027-12-31 humanoid actuators consume less than 1% of annual global sintered NdFeB output on any published third-party accounting, and no humanoid platform in series production discloses a rare-earth-free primary joint actuator before 2028-12-31.

HELD open (three lenses agree). Plum Acquisition Corp IV / CTR EX-99.1 decks (SEC, 2026-01-28, 2026-08-20; verified in the local EDGAR corpus): 'Demonstration Plant operating at 1/15th commercial scale', footnote 'Commercial scale defined as 1/15th of production scale', 770 h campaign; Stage 1 (25 kt/yr LiOH) contingent on a H2-2026 SPAC close at $4.7bn. 'First full year of continuous brine processing' is found only in promotional aggregators and is not admitted.

★ For the kids: Very hot salty water under the California desert has lithium in it. A small test plant pulled some out; 'proven' here means 'tested small'.

Technical: Salton Sea brine ~200 mg/L Li, >250 g/L TDS, high Fe/Mn/Zn/silica, ~250 C; silica and iron management precede any sorbent. Resource base: LBNL 2023 (~18 Mt LCE probable).

Frozen prediction: No Salton Sea or Smackover DLE plant produces >=5,000 t LCE in any 12-month window ending on or before 2027-12-31 (p=0.8); by 2028-12-31 CTR Stage 1 has not reported first LiOH production (p=0.75).

MERGED (lepidolite-mica-lithium + lepidolite-lithium-swing-supply) and held VIABLE on the production fact alone: mica-hosted lithium is commercially converting at scale and is the price-setting marginal tonne. CATL's Jianxiawo (~46 kt/yr LCE-equivalent, ~3% of world output) suspended 2025-08-10 on a purely administrative permit expiry, stayed down about eleven months, and restarted early July 2026 on a licence running only to 2028-02-27; GFEX lithium carbonate futures more than doubled, topping 200,000 yuan/t. The MARGINAL-COST-SETTING sub-claim is held at modeled — a price move around one ou…

★ For the kids: A lot of China's lithium comes from a flaky rock with only a little lithium in it. When one mine's paperwork expired, the world lithium price doubled.

Technical: Lepidolite is ~1-2% Li2O in ore versus 6% in spodumene concentrate, carries fluorine, Rb and Cs, and yields 8-20 t of slag per tonne LCE. Scale figures are TRADE-PRESS grade (Benchmark Source, Fastmarkets, Aug 2025) and are not treated as audited. Consequence for the ledger: DLE, recycling and sodium-ion are all graded against a lithium price whose short-horizon variance is dominated by Chinese permitting.

Frozen prediction: At least one further Chinese lepidolite-linked suspension or permit-driven curtailment of >=20 kt/yr LCE-equivalent is publicly reported before 2027-12-31; and USGS MCS 2028 does not report Chinese lithium MINE production below its 2025 level.

HELD modeled. Ascend Elements Chapter 11 (2026-04-10) after Li-Cycle (May 2025) = count 2; the standing >=10 kt/yr nameplate clause is 'true pending nameplate verification' from first-day declarations, not asserted. Lithium carbonate ~148,000 RMB/t (Aug 2026) is the countervailing variable. Promotion to viable needs a third insolvency or an audited utilisation figure.

★ For the kids: Battery recycling factories were built before enough old batteries existed to feed them. A second big American recycler ran out of money in 2026.

Technical: Hydromet plants need >10 kt/yr black mass (Baum et al. 2022); EV end-of-life volumes are ~5 years behind Harper et al. 2019; grant-dependent capex stacks fail first; EU 2023/1542 recycled-content floor binds from 2031.

Frozen prediction: Standing: >=1 more Western Li-ion recycler with >=10 kt/yr nameplate enters insolvency/idling/distressed sale by 2027-12-31 (Ascend satisfies it if Covington nameplate >=10 kt/yr is verified); NEW: by 2027-12-31 no Western black-mass hydromet plant discloses audited utilisation >=60% of nameplate (P=0.7).

HELD modeled. Rotterdam 99.65% antimony ~$23,000/t July 2026 (second monthly fall from the ~$59,750/t 2025 peak); Perpetua Stibnite $2.9bn EXIM approval (May 2026, ~2029 production); USAC Thompson Falls restart (~576 t ore). 2026-11-27 suspension expiry is the next gate. Pantry lead 693489448f4a (antimony electrowinning) was not found in the local corpus and is correctly dropped.

★ For the kids: Antimony got very expensive when China stopped selling it to the US; now the price is coming down even though the new American mine will not open until about 2029.

Technical: ~85% by-product/co-product supply; Western primary supply before 2029 limited to small stibnite hauls and toll smelting; price decline = demand destruction plus licensed Chinese flow.

Frozen prediction: Standing: suspension lapses or extends on 2026-11-27 without a return to blanket ban and Rotterdam monthly average stays <$40,000/t through 2026-12-31; NEW: no US primary antimony mine ships >=1,000 t contained Sb in any 12-month window ending on or before 2028-12-31 (P=0.75).

HELD viable: Antofagasta-Jiangxi 2026 benchmark TC/RC settled at $0/t and 0 c/lb (Reuters, 2025-12-19), from $21.25/t and 2.125 c/lb in 2025 - a verified structural fact, not a forecast.

★ For the kids: Copper factories normally get paid to turn rock into metal. This year they agreed to do it for free because there is not enough rock to go around.

Technical: At zero TC/RC, custom-smelter margin = sulphuric acid + precious-metal credits + free-metal terms.

Frozen prediction: By 2027-03-31 at least one non-Chinese custom copper smelter announces permanent closure or indefinite care-and-maintenance citing TC/RC economics (P=0.65).

HELD at open, and the four-decade gap is itself the finding. Billions of tonnes stockpiled with ~150 Mt/yr added, demonstrably containing scandium, rare earths, gallium, titanium and iron. The laboratory literature is strong and reproducible — pilot-plant scandium leaching published in 2002, and ~75% Sc recovery with ~88% of contained REEs isolated as a by-product in 2018. What does not exist anywhere after twenty-four years is a commercial plant. The April 2026 US Critical Materials / Columbia 'Mud To Metal' agreement is a two-year research programme and cannot produce one inside the window.

★ For the kids: Making aluminium leaves billions of tonnes of red mud with valuable metals in it. Scientists have proved for forty years they can get the scandium out. Nobody has ever built a factory that makes money doing it.

Technical: Economics are dominated by grade and matrix, not recovery chemistry: Sc runs 100-150 ppm and is locked in hematite/goethite, so selective dissolution means dissolving iron you must re-reject, while residual Bayer alkalinity (pH 10-13) consumes acid before any target metal dissolves. Every proposed flowsheet moves the problem rather than removing it. The credible path is co-recovery of TiO2 and iron oxide so the mass burden becomes a product.

Frozen prediction: No facility outside China producing >=100 t/yr contained REO or >=20 t/yr Sc2O3 from bauxite residue is operating by 2028-12-31.

HELD at open, with the performance claim struck. KoBold applied ML geological modelling after acquiring Mingomba in December 2022 and broke ground on construction in May 2026 targeting ~300 kt/yr copper at ~1,700 m for $2.3-2.5bn. The promotional 4-5x compression claim fails three ways: n=1; the Zambian Copperbelt was among the most intensively explored terrains on Earth, so 'the model found what geologists missed' is not separable from 'someone finally paid to drill deep in a known district'; and by May 2026 the project was reported hitting hard ground technically and financially.

★ For the kids: A computer learns which patterns mean metal is underground, then picks where to drill. It found copper in Zambia — but people had searched there for a hundred years, so we cannot tell if the computer was clever.

Technical: The method is well-posed and unremarkable as statistics: joint inversion and supervised learning over co-registered geophysics, geochemistry, drill logs and structure to rank targets; the genuine advantage is a portfolio effect. The grading problem is structural — ground truth is drilling, you only observe outcomes for targets you chose to drill, so precision is measurable while recall is not. The honest test is pre-registration.

Frozen prediction: No ML-exploration operator publishes a pre-registered ranked drill-target list with an audited hit rate against a conventional baseline before 2028-12-31.

HELD on the two-axis rule. CdTe modules are unambiguously commercial, but the CHOKEPOINT claim rests on price and a licensing regime rather than a documented physical allocation failure. More than 90% of tellurium comes from primary electrolytic copper anode slimes; 70% of end use is solar; China refines ~80% (800 t of a ~1,000 t world total) and has licensed exports since 2025-02-04; the 2025 annual average rose 84% in Europe to ~$150/kg and 60% in the US to ~$120/kg. The US refines none.

★ For the kids: One kind of solar panel needs tellurium, which only ever turns up as a leftover from making copper. One country cleans up most of it, so when exports slowed the price nearly doubled.

Technical: USGS MCS 2026 (Tellurium, Feb 2026). Control instrument: MOFCOM/GAC Announcement No. 10 [2025], effective 2025-02-04, 41 HS codes spanning tungsten, tellurium, bismuth, molybdenum and indium; licences take ~45 days and lapse on any change of recipient or end use. Read jointly with chalcopyrite-catalytic-leaching: relief in copper tightens this.

Frozen prediction: USGS MCS 2028 still reports China at >=70% of world refined tellurium production AND US tellurium refinery production of zero.

HELD. Secondary recovery from ITO sputter scrap in Japan and Korea is commercial; ITO SUBSTITUTION is not — USGS has listed the same six substitute classes (antimony tin oxide, CNT films, PEDOT, Cu/Ag nanowires, graphene, ZnO nanopowder) for years while ITO remains the majority use. China refines 70% (760 t of 1,100 t, 2025); the US refines none and is 100% import-reliant on ~220 t/yr. Unwrought-indium exports fell 72% year-on-year to September 2025 after the February 2025 controls.

★ For the kids: Touchscreens need a see-through metal coating made from indium. Nobody mines it on purpose, so factories recycle scrap instead of finding a replacement.

Technical: USGS MCS 2026 (Indium). PRIMARY-SOURCE INCONSISTENCY the engine must not launder: the salient-statistics table gives a 2025 US warehouse annual average of $370/kg while the narrative says $390/kg (June peak $408/kg). The TABLE value, US warehouse, fob, 99.99% minimum purity, is the only series frozen here; any citation not naming the series is inadmissible.

Frozen prediction: Every USGS MCS edition through the 2029 volume continues to state that ITO accounts for most global indium consumption AND reports US primary indium refinery production of zero.

HELD — the chokepoint is well documented while every proposed escape is pilot-scale. Roughly 70-90% of world HPQ comes from two mines around Spruce Pine, North Carolina, feeding the fused-quartz crucibles used to pull every Czochralski silicon ingot for both semiconductor wafers and PV. Synthetic crucibles are purer but cost roughly 5-10x; the two Chinese deposits announced in April 2025 have pilot output at 4N8 (99.998%), below the ultra-high-purity inner-layer requirement of <=10 ppm total impurities, and crucible qualification runs years.

★ For the kids: Almost all the super-clean sand used to melt computer-chip silicon comes from two mines in one small town. The substitutes cost many times more and the new mines' sand is not clean enough yet.

Technical: USGS MCS 2026: HPQ <100 ppm total impurities (>=99.99% SiO2); ultra-high-purity <=10 ppm. Beneficiation evidence that non-Spruce-Pine feedstock is hard: Tang, Sun, Zhang & Peng, Miner. Eng. (2026), DOI 10.1016/j.mineng.2026.110191 — note the target there is the looser PV crucible spec. Fluid inclusions and lattice-substituted Al are the impurities acid leaching and chlorination cannot fully remove. Reaches the telescope lobe too: the same feedstock is upstream of fused-silica optical blanks.

Frozen prediction: As of 2028-12-31 no Chinese Henan- or Xinjiang-sourced HPQ is qualified into the crucible inner-layer specification of a top-five silicon wafer manufacturer, and Spruce Pine still supplies the majority of world ultra-high-purity crucible-grade quartz.

HELD, and the marketing claim is partly false. Chalcogenide glasses are genuinely commercial for MWIR and LWIR lenses and are precision glass-mouldable, which is the real reason they displace diamond-turned germanium. But the workhorse commercial compositions are Ge-BEARING: AMTIR-1 is Ge33As12Se55 and the GASIR family is Ge-As-Se. These THRIFT germanium, they do not eliminate it. Genuinely Ge-free options (As2Se3, As2S3, Sb-Se) carry worse dn/dT, lower Tg, softer mechanics and arsenic-handling burdens.

★ For the kids: Heat cameras normally use lenses made from germanium. New glass lenses replace some of it — but a lot of that replacement glass still has germanium inside, about a third by weight.

Technical: Matsushita, Sato, Masuda & Kimata, Opt. Eng. 62(8), 087104 (2023); Zhu, Luo, Zhou & Li, Chin. Opt. Lett. 24, 072201 (2026). The grading hook is compositional, so it is falsifiable from supplier datasheets rather than press releases. Transferable lesson: the material that wins is usually the one with the better forming process, and supply-risk narratives get attached afterwards.

Frozen prediction: Through 2028-12-31 the majority of chalcogenide glass sold for LWIR imaging optics remains germanium-bearing rather than Ge-free.

HELD. VRFBs are pitched as escaping lithium's supply risk; they substitute one concentrated dependency for another. China is roughly 70 kt of 2024 vanadium production (~70% of world), Russia ~21 kt, South Africa ~8 kt — a more concentrated top three than lithium. Electrolyte is 30-50% of VRFB capex at roughly $140-180/kWh in 2026, so the vanadium price is the chemistry's cost derivative. The genuine structural mitigant is that vanadium is not consumed: electrolyte leasing converts a capex problem into a financing problem, which is a real advance but a financial one.

★ For the kids: Some giant grid batteries use vanadium instead of lithium. Most vanadium comes from three countries and the liquid is about half the cost, so it is not really escaping the problem.

Technical: USGS MCS 2025/2026 for production by country. V2O5 fell from ~$9/lb (2022) to $4-5/lb (2024) and traded around $27-29/kg in Europe through H1 2026 — a trade-press/consultancy series, not exchange-settled, and treated as indicative only. Electrolyte is V3.5+ in sulfuric acid; the 3.5+ preparation route is itself a cost lever. VRFB deployment is therefore sensitive to the cost of capital in a way lithium storage is not.

Frozen prediction: Vanadium electrolyte still exceeds 25% of installed VRFB system capex in credible published 2028 cost breakdowns, and no non-vanadium flow chemistry exceeds 20% of annual global flow-battery MWh deployments before 2028-12-31.

HELD. Every PEM electrolyser and automotive fuel cell runs on a perfluorosulfonic-acid membrane (Chemours Nafion, Syensqo Aquivion, AGC, Gore). The membranes ship commercially today; what is modelled is the RESTRICTION exposure, from two datable pressures: 3M's 2022-12-20 announcement that it would exit all PFAS manufacturing including Dyneon fluoropolymers by end-2025, and ECHA RAC's final opinion (2026-03-02) keeping fluoropolymers inside the universal PFAS restriction.

★ For the kids: The heart of a hydrogen machine is a thin plastic sheet made with forever chemicals. Rules may restrict them, and one big maker already stopped.

Technical: Kusoglu & Weber, Chem. Rev. 117, 987 (2017): no non-fluorinated membrane matches PFSA on combined conductivity, gas crossover and chemical stability at electrolyser duty. Score JOINTLY with pfas-fluorine-battery-materials and fluorspar-hf-fluorochemical-supply — same fluorine chain, and netting them silently hides both. Iridium is the CATALYST chokepoint; PFSA is the MEMBRANE chokepoint; a stack needs both solved.

Frozen prediction: No commercially sold PEM electrolyser or automotive fuel-cell stack ships a non-fluorinated membrane as its primary ion conductor at >=10 MW cumulative shipped capacity before 2029-12-31.

HELD at open. AEM is the only route attacking both tracked hydrogen chokepoints at once — non-noble catalysts (no iridium) on a hydrocarbon membrane (no PFSA) — and it is the least proven. Small commercial units exist; multi-megawatt field durability does not. Vendor stack-life EXPECTATIONS (>=35,000 h) are excluded from tier support; the peer-reviewed durability frontier is roughly 1,000 h at 1 A/cm2, more than an order of magnitude short.

★ For the kids: A newer hydrogen machine skips both the very rare metal and the forever-chemical plastic. It works small; nobody has shown it lasts big.

Technical: Miller, Bouzek, Hnat et al., Sustainable Energy Fuels 4, 2114 (2020): blockers are membrane and ionomer alkaline stability and carbonation, not activity — the same durability-not-activity pattern as PGM-free fuel-cell catalysts. Promotion gate frozen before the fact: a >=10 MW single-site installation in commercial operation, an INDEPENDENT non-vendor report of >=20,000 h at >=1 A/cm2, and third-party-verified zero iridium content.

Frozen prediction: No AEM installation of >=10 MW at a single site enters commercial operation, and no independent non-vendor report documents >=20,000 h of AEM stack operation at >=1 A/cm2, before 2028-12-31.

HELD at open, and it cuts AGAINST the thrifting story. Certified cells reach ~34.9% with ~30.1% at module level; Qcells has certified M10-format tandem cells at 28.6% from a pilot line; Oxford PV made the first commercial tandem module shipment in September 2024. All of it is pilot-line volume with no multi-year field-degradation record. A certified champion cell is not a tier, and a first commercial shipment is not a commercial technology. The materials read that matters: tandems and HJT need ITO/IZO transparent electrodes, so an efficiency transition RAISES indium intensity per watt while s…

★ For the kids: A new solar cell stacks two light-catching layers. It wins records in the lab, but nobody knows if it survives twenty years on a roof.

Technical: Al-Ashouri et al., Science 370, 1300 (2020). Stability discipline: Khenkin et al., Nature Energy 5, 35 (2020) — the ISOS consensus protocols exist precisely because stability claims without specified protocols are not comparable. Grade jointly with silver-thrifting-pv-metallization and indium-supply-ito-thrifting: 'PV is dematerialising' is only ever true metal by metal.

Frozen prediction: Through 2028-12-31 perovskite-silicon tandem modules stay below 1% of annual global PV module shipments, AND no tandem product carries a 25-year power warranty backed by >=5 years of published outdoor field data.

DEMOTED modeled -> open. Mo/Si multilayers for EUV lithography are commercially shipping and W/Si flew on NuSTAR — but those are settled matters outside this record's distinguishing claim. What this component actually tracks is laterally graded Cr/Sc and W/B4C multilayers for soft X-ray SPECTROPOLARIMETRY, and the sole evidence is one non-peer-reviewed arXiv preprint of a single proof-of-concept (Cr/Sc near Brewster's angle over 200-400 eV; W/B4C at 30 degrees over 500-850 eV). A preprint cannot carry 'modeled'. The export-control facts are documented and stand on their own: scandium was adde…

★ For the kids: X-ray space telescopes use mirrors made of hundreds of paper-thin metal layers. The newest ones use scandium and tungsten, two metals now hard to export from China.

Technical: Heine, S., Marshall, H. L., Garner, A., Gullikson, E. & Kothnur, N., arXiv:2607.27664, submitted 2026-07-30 — cited as a frontier marker only, carrying no maturity weight, and verified at source from pantry lead claim:kraken:a97a0fd0f3b4. Established comparators: Cr/B4C interface work, J. Appl. Phys. 119, 125307 (2016); V/Sc water-window multilayers, Sci. Rep. 7, 12081 (2017). Note the real exposure is sputtering-target PURITY and qualified vendor count, not tonnage — scandium mass per mirror set is negligible against world Sc2O3 output.

Frozen prediction: Scandium and/or tungsten sourcing is named as a schedule or cost risk in at least one publicly documented X-ray or EUV optics programme before 2028-12-31; and no independent peer-reviewed replication of laterally graded Cr/Sc spectropolarimetry performance appears before 2028-12-31.

NEW, accepted at modeled. LG Chem, Samsung SDI/EcoPro and L&F position single-crystal mid-Ni NMC charged to 4.35-4.4 V as the cost bridge between LFP and Ni-rich (single-crystal tolerance literature: Li, Dahn et al., J. Electrochem. Soc. 2017-2019). Vehicle programs are announced for 2025-2027; no volumes reported.

★ For the kids: Instead of packing in more nickel, engineers squeeze more energy from a medium-nickel battery by charging it a little harder.

Technical: Raising cut-off 4.2 -> 4.4 V on Ni0.65Mn0.2Co0.15 recovers ~10% capacity; needs single-crystal morphology, LiFSI/fluorinated additives and coatings against oxygen release and TM dissolution.

Frozen prediction: By 2027-12-31 at least one OEM discloses a series-production EV using NMC with Ni<=0.7 charged to >=4.35 V.

NEW, accepted at open. Phosphate/phosphazene non-flammable electrolytes have long lab records (Wang et al., Nature Energy 2018) but commercial cells use only a few wt% of additive because phosphates degrade graphite SEI and rate. China GB 38031-2025 no-fire/no-explosion rule (effective 2026) creates pull; no evidence of >10 wt% adoption.

★ For the kids: Battery liquid can burn, so chemists add fire-resistant ingredients, but too much makes the battery slow.

Technical: Phosphates co-intercalate into graphite; mitigations (LHCE with TTE, FEC-derived SEI) reintroduce fluorinated/PFAS exposure.

Frozen prediction: No top-10 cell maker discloses a mass-produced EV cell with >=10 wt% phosphate/phosphazene flame retardant by 2027-12-31.

NEW, accepted at modeled. FEC demand scales with silicon-anode adoption (5-10 wt% in Si-C vs 1-2 wt% graphite); production concentrated in China (Tinci, Capchem, Huasheng) with Japanese/Korean minority supply. Concentration is documented in supplier filings; the 2021 VC price spike is the only realised stress event, so modeled.

★ For the kids: A pinch of special ingredients makes the battery skin form properly; most of the world's pinch comes from a few factories in China.

Technical: FEC via fluorination of ethylene carbonate (HF-linked; fluorspar chain); LiDFP from LiPF6 routes; FEC decomposition raises gassing.

Frozen prediction: If silicon-dominant EV anodes reach >5% of EV cells by 2028, FEC spot prices in China show at least one >2x spike within 12 months.

HELD modeled (conservative; the regulation itself is a shipping fact, the compliance-supply thesis is not). Pantry lead 71d6be28e80e traced to EMAT EX-99.1 (2026-08-10; corpus lines 2334/2349): 'only known company outside China' claim, ULVAC furnace order, 10,000 t target - all excluded evidence classes, and contradicted by MP Fort Worth shipping since Jan 2025. Cross-link: magnet-provenance-forensics shows compliance is documentary, not verifiable.

★ For the kids: From next year the US military may not buy magnets made from Chinese rare earths. Several companies say they are ready; only one has actually shipped magnets.

Technical: From 2027-01-01 NdFeB and SmCo magnets with REE mined/refined/separated/melted in China, Russia, DPRK or Iran are barred from DoD end items; compliance via attestation and audit.

Frozen prediction: By 2027-06-30 EMAT's 10-K/10-Q cumulative finished-NdFeB magnet revenue is below $10M (P=0.8).

ACCEPTED at viable ONLY for the zinc-tailings branch: Century tailings (Queensland) hydraulic re-mining shipped zinc concentrate commercially 2018-2024 (New Century Resources ASX; Sibanye-Stillwater annual reports 2023-24). Jetti catalytic re-leach at Pinto Valley (cathode since 2019) is a single-site commercial branch = modeled. Cobalt-from-tailings, monazite tails and bauxite-residue REE = open. The viable tag must not be read across branches.

★ For the kids: Old mines left hills of crushed rock that still have metal in them. At one huge zinc mine in Australia, machines washed metal out of those leftovers and sold it for years.

Technical: Crushing/grinding is sunk; economics depend on liberated mineralogy and water balance; Jetti's catalyst targets chalcopyrite passivation.

Frozen prediction: Through 2027-12-31 no cobalt-from-tailings project in North America ships >=500 t contained cobalt.

HELD open. FBG / Fabry-Perot / IR-fibre operando sensing demonstrated in commercial 18650 cells (Huang et al., Nature Energy 2020; Gervillie-Mouravieff et al., Nature Energy 2022; lab-on-fibre thermal-runaway precursors, Nature Communications 2023). Single-cell lab formats only; zero mass-produced cells ship with embedded fibre; hermetic feedthrough and BMS integration unsolved.

★ For the kids: Tiny glass threads inside a battery that let it feel its own temperature and smell its own chemistry before anything goes wrong.

Technical: FBG (T/strain), Fabry-Perot (pressure), chalcogenide IR fibre for SEI/solvent signatures; not module-level.

Frozen prediction: By 2028-12-31 no top-10 cell maker ships a mass-produced cell with an embedded optical fibre (P=0.85).

DEMOTED modeled -> open. Ultrasonic motors are a shipping niche technology (Canon USM; Uchino 1997), but the COMPONENT CLAIM is NdFeB demand displacement, and for that no model, tonnage estimate or OEM disclosure exists. 'Modeled' requires a model; 'a mature technology plus an unquantified thesis' is open. Trade-off noted: PZT is lead-based (RoHS exemption), lead-free KNN (Saito et al., Nature 2004) has not displaced it.

★ For the kids: Some tiny motors, like a phone camera's focus, can use a crystal that wiggles with electricity, so no magnet is needed.

Technical: Small magnet motors are unit-heavy, tonnage-light; watch for any OEM disclosure of NdFeB tonnage avoided.

Frozen prediction: P=0.7 that no published supply-demand model (Adamas, USGS, IEA) attributes >=1% of NdFeB demand displacement to piezo/ultrasonic substitution by 2028-12-31.

DEMOTED modeled -> open. YASA axial-flux motors are in production (Mercedes-AMG), but the component claim is reduced NdFeB kg/kW, and no OEM or teardown publishes magnet mass per kW by topology. A production motor with an unmeasured claim is open; promote to modeled on the first independent teardown with magnet mass and peak kW.

★ For the kids: This motor is shaped like a flat pancake instead of a ring around a rod; it can be lighter, so it might need less magnet - nobody has weighed it yet.

Technical: Yokeless segmented armature; ~50 kW/kg headline; magnet mass undisclosed. Grading source: Munro / IDTechEx / A2Mac1 teardown.

Frozen prediction: P=0.5 that an independent teardown publishes NdFeB kg/kW for a production axial-flux EV motor by 2027-12-31.

HELD viable as a shipping policy fact. Suspension of Announcements 55/56/57/58/61/62 runs to 2026-11-10; the April 2025 licence regime on Sm/Gd/Tb/Dy/Lu/Sc/Y never lapsed. Dated watch item nine weeks out.

★ For the kids: China paused its toughest magnet export rules for one year; the pause ends 10 November 2026.

Technical: April list covers medium/heavy REE, not NdPr; HRE-free and GBD-thrifted grades sit partly outside scope.

Frozen prediction: P=0.4 the suspended October 2025 extraterritorial magnet rules are reinstated in whole or part by 2027-06-30; count watch: number of Announcements extended/reinstated/replaced on or before 2026-11-10.

HELD open (spectral loner). Sulfate-route hydromet (black-mass leach, pCAM coprecipitation, HPAL, REE sulfation) produces ~1.5-2.5 t Na2SO4 per t product; documented in process models (Argonne EverBatt; Harper et al., Nature 2019; Baum et al., ACS Energy Lett. 2022) but relief routes (BMED acid/base splitting, sellable-grade crystallisation, chloride routes) are pilot or paper. Falsifier stated in the study.

★ For the kids: Pulling metals out of old batteries uses acid, then base to cancel it. The leftover is a lot of salt, and nobody has a great place to put it yet.

Technical: MSO4 + 2 NaOH -> M(OH)2 + Na2SO4; BMED ~1.5-2.5 kWh/kg NaOH-eq at pilot; membrane fouling by ppm Ni/Co/Mn is the open engineering problem.

Frozen prediction: By 2027-12-31 at least one EU or Korean recycling/pCAM plant commissions a commercial (not pilot) Na2SO4 splitting or sellable-grade crystallisation line (P=0.35).

HELD modeled. Supply concentrated (Qatar/US/Russia-Amur; USGS MCS 2025); Federal Helium System sold to Messer (BLM 2024). Shipping relief: dry dilution refrigerators/pulse-tube cryocoolers (>15 yr commercial) and sealed MRI magnets (Philips BlueSeal ~7 L). Not viable overall: fab/preform demand still vents helium and fab recovery is pilot-to-early-commercial.

★ For the kids: Helium is balloon gas, but it also keeps hospital scanners and chip factories cold. Once it floats away it leaves Earth forever, so the smart trick is machines that keep it in a loop.

Technical: Membrane + PSA/cryogenic upgrading of vent gas; PT cryocooler base 2.5-4 K; sealed MRI fill <10 L vs ~1,500 L.

Frozen prediction: By 2027-12-31 no new ex-Qatar/ex-US/ex-Russia helium source >200 MMcf/yr enters production (P=0.7).

HELD modeled. DNANF loss 0.091 dB/km (Southampton ORC / Microsoft, Nature Photonics 2025); Microsoft/Lumenisity reports >1,000 km in Azure routes. Single vendor, bespoke splicing/connectorisation, lab-spool loss not installed-route loss: pilot-commercial, not viable. Vendors with >100 km installed = 1.

★ For the kids: Normal fibre sends light through glass; this one sends it through air inside a glass straw, so it goes faster and loses less.

Technical: Nested anti-resonant silica tubes; ~47% faster group velocity than SMF-28; ~1,000x lower nonlinearity; 1.3-1.6 um window.

Frozen prediction: By 2028-06-30 a second manufacturer ships >100 km of hollow-core fibre for a terrestrial telecom route (P=0.4, count-based watch item).

HELD open. Electra's 500 t/yr Colorado demonstration plant guided mid-2026, still under construction Aug 2026 ('by end of 2026'); $186M raise and offtake-type commitments are excluded evidence.

★ For the kids: Instead of a giant furnace, this plant uses electricity to pull iron out of a liquid. The first small factory is still being built and keeps getting delayed.

Technical: Acid dissolution ~60 C, impurity separation, electrowinning of 99% Fe plate; metric = tonnes of plate/month vs 500 t/yr design, then current efficiency and kWh/t.

Frozen prediction: Standing: by 2027-12-31 no electrochemical-iron plant reports >5,000 t/yr actual output (P=0.85); NEW: Electra does not report first plate iron before 2027-03-31 (P=0.5).

SKEPTIC: DEMOTED viable -> modeled on the split-tier rule this map already applies to samarium-iron-nitride-magnets (bonded viable / sintered open => record held at modeled). The XRT branch on coarse tungsten/tin/diamond ore is genuinely commercial (Mittersill in production since 2012; Robben & Wotruba, Minerals 2019), but the record as slugged spans XRT (commercial), NIR-on-spodumene (vendor case studies, no audited recovery) and LIBS/hyperspectral (open), and its own prediction is a p=0.5 count-based watch admitting no audited ex-China lithium circuit. A family record cannot carry viable on…

★ For the kids: Before grinding rocks to get the metal out, a machine looks through each rock with X-rays and throws away the empty ones. It works for tungsten and diamonds; for lithium rock it is still being tested.

Technical: XRT resolves atomic-density contrast per particle at up to ~150 t/h per metre belt width; NIR resolves mineralogy by reflectance; LIBS gives elemental spectra at particle-level throughput limits. Admissible evidence = the audited pair (mass rejected %, metal lost %) from a technical report.

Frozen prediction: By 2028-12-31 at least one ex-China lithium or tungsten concentrator discloses in an NI 43-101 / JORC / issuer filing an operating sensor-based sorting circuit with measured mass rejection >=20% and metal loss <=5% (p=0.5).

HELD open. Ah-level pouch results at -40 to -50 C are single-lab peer-reviewed: 1.2 Ah Na-ion pouch, 163 Wh/kg (25 C) / 107 Wh/kg (-50 C), 93% retention over 200 cycles at -50 C (Li et al., Nature Communications, Feb 2026); weakly-solvating ester/ether designs give reversible Li cycling at -40 C (Nature Communications 2025; Advanced Energy Materials 2025 reviews). CATL's Naxtra >90% at -40 C is a company claim. No Tier-1 maker names such an electrolyte in a shipping-cell datasheet.

★ For the kids: Batteries get sluggish in the cold because charged particles must shed a coat of solvent before entering the electrode. Chemists are making thinner coats so they slip in even at -50 C.

Technical: Design lever = weaken cation-solvent binding (low donor-number esters/nitriles, sterically hindered solvents) or bypass desolvation via solvent co-intercalation into hard carbon; trade-off is lower ionic conductivity and salt solubility.

Frozen prediction: By 2027-12-31 a Tier-1 cell maker names a weakly-solvating or ester-based low-temperature electrolyte in a shipping-product spec or peer-reviewed paper on a >1 Ah production cell. p=0.30.

HELD open. Weng et al. (Joule 2021) and a Stanford data-driven study (Joule 2024) link formation protocol to lifetime and electrode utilization; Fan et al. (Nature, June 2026) show 2 C initial charge on Li-rich oxides gives +20% capacity and >36% cycle life via residual-Li self-pinning. All coin/small-pouch; no gigafactory adoption published.

★ For the kids: A new battery needs a few careful first charges before it is sold, like breaking in shoes. For some cathodes a quick break-in works better and saves factory time.

Technical: Mechanism differs by chemistry (graphite/NMC: SEI porosity and Li inventory; Li-rich: residual Li suppresses O-redox lattice collapse). Diagnostics enabling adoption: end-of-formation dQ/dV and low-SOC resistance.

Frozen prediction: By 2027-12-31 at least one cell manufacturer or automaker publicly states a >=30% formation-time reduction attributed to a new protocol on a production line. p=0.35.

HELD open (new domain entered at the bottom tier, as required). Bothamy & Galy (Front. Environ. Chem. 2021) and Suarez-Criado et al. (JAAS 41, 1675, 2026) both disclaim geographic origin attribution; ISO 17887:2025 and ISO 23664:2021 are records standards; DFARS 252.225-7052 (Federal Register 2024-11513; corpus lead 71d6be28e80e verified) restricts from 2027-01-01 on stages no fingerprint can resolve. No blind-classification study exists.

★ For the kids: Rules now say magnets must not come from certain countries, but no lab test can tell where a magnet's metal was born - only the paperwork says so.

Technical: Mass-dependent Nd fractionation flags processed Nd; radiogenic 143Nd/144Nd is erased by feed blending and swarf loops. A usable tool needs a blind test: >=50 finished magnets, >=3 separation countries, reported accuracy.

Frozen prediction: P=0.75 that by 2028-12-31 no peer-reviewed study blind-classifies finished sintered NdFeB by country of separation at >=90% accuracy on >=50 samples from >=3 countries; P=0.7 that through 2027-12-31 DoD adds no analytical-verification requirement to 7052 compliance.

HELD open. The only public interlaboratory comparison (NPL Report MAT 54, 2012, IEC TC 68 / NIM Beijing) found a significant HcJ method dependence attributed to magnetic viscosity; zero comparisons found 2013-2026. Consequence: every 'HcJ >= X' bar in this map is method-conditional.

★ For the kids: Two labs can measure the same magnet and get different 'strength' numbers because one measures fast and one slow; nobody has re-checked the gap since 2012.

Technical: PFM applies 5-10 T in ms; viscous reversal (S ~1-2%/decade) shifts apparent HcJ up vs quasi-static closed-circuit; eddy currents and open-circuit corrections (IEC 60404-7) compound the spread for <1 g samples. To modeled: a published multi-lab comparison on >=1 HRE-free grade with uncertainty budgets.

Frozen prediction: P=0.7 that no interlaboratory comparison (>=5 labs) on sintered NdFeB HcJ using both IEC 60404-5 and pulsed-field methods is published by an NMI or IEC TC 68 by 2028-12-31.

HELD open. Laboratory rigs and retrofit machines only (Suryandi et al., Machines 2022 review; e-Prime 2025 dual flux/temperature Terfenol-D FBG); zero production traction motors with embedded fibre; no OEM attributes any Dy/Tb reduction to sensing.

★ For the kids: A glass thread inside the motor could tell it how hot its magnets really are, so engineers would not have to add extra rare earth 'just in case'.

Technical: FBG ~10 pm/K, ~1.2 pm/microstrain at 1550 nm; Terfenol-D magnetostriction is nonlinear, hysteretic and temperature cross-sensitive; rotor-side sensing needs rotating interrogation.

Frozen prediction: P=0.85 that no series-production passenger-EV traction motor (>10,000 units/yr) ships with an embedded fibre-optic magnet flux or rotor-temperature sensor by 2028-12-31; P=0.8 that no OEM/Tier-1 attributes a Dy/Tb reduction to in-situ sensing in that window.

HELD open. Kaarlela et al. (J. Manuf. Syst. 74, 2024) systematic review: fully automatic disassembly 'remains difficult'; ORNL/CMI 2021 and Free4Lib demonstrators are lab-scale; zero commercial no-human-in-loop lines.

★ For the kids: Taking apart an old EV battery is a puzzle where every car company uses different screws and glue. Robots can do it in the lab; no factory lets robots do it alone yet.

Technical: Blockers: pack design variability, adhesive/potted modules, HV and thermal-runaway safety during cutting, vision under occlusion. Mechanical precondition for direct-cathode recycling.

Frozen prediction: By 2028-12-31 no Western recycler publicly reports a fully automated pack-to-cell disassembly line processing >=10,000 EV packs/yr (P=0.85).

HELD modeled. Taseko Florence first cathode late Feb 2026; Q1 1.5 Mlb, Q2 5.2 Mlb (~25% of 85 Mlb/yr annualised), 2026 guidance 30-35 Mlb. One site, one operator, FOAK well-field ramp; oxide-only, no bearing on the chalcopyrite chokepoint. Note the lens's 'first greenfield US copper since 2008' is a press framing; copper ISR itself has prior commercial history (San Manuel), so novelty is not a tier input either way.

★ For the kids: Instead of digging a huge pit, they pump weak acid down wells, it dissolves the copper, and they pump the copper water back up. It started in Arizona in 2026.

Technical: Raffinate injection, PLS recovery, SX/EW to LME Grade A; gradeable variables are quarterly cathode pounds vs nameplate and PLS grade as wells come on; aquifer hydraulic control is the licence risk.

Frozen prediction: Florence does NOT reach >=60 Mlb of cathode in any rolling 12-month window ending on or before 2027-12-31 (P=0.6); if it does, promote to viable.

HELD viable, on the narrow reading that what is replication-grade is the GAP itself: three independent groups (Chen et al., Joule 2019; Randau et al., Nature Energy 2020; Puls et al., Nature Energy 2024, 21-lab round robin) converge on coin-cell numbers not transferring to practical cells. No fix is viable; the viable tag licenses only the filter rule: a cell figure without E/C ratio, N/P ratio, areal capacity, stack pressure and format cannot support a tier above open.

★ For the kids: When scientists test a tiny toy battery it often looks amazing, but the real-sized one loses the magic. Labs are writing rules so everyone measures the same way.

Technical: Chen: coin-cell life dominated by flooded electrolyte (>>3 g/Ah) and thick Li; Randau: minimal Li | beta-Li3PS4 | NMC622 reference cell; Puls: same materials and protocol, own assembly -> wide spread.

Frozen prediction: By 2027-12-31 fewer than 5 of the 10 most-cited 2026 primary papers on Li-metal or ASSB anodes in Nature/Science/Nature Energy/Joule report E/C, N/P, areal capacity and stack pressure together in the main text (P=0.7, reviewer count).

HELD viable: commercially shipping for decades and replicated at series scale - SCHOTT delivered all 949 ZERODUR ELT M1 blanks 2019-2024 (SPIE 12188, 2022; ESO eso2410). Three-vendor know-how oligopoly (SCHOTT, Corning, Ohara); consistent with the 07-31 kill sic-displaced-glass-ceramic-mirrors.

★ For the kids: Giant telescope mirrors are made of a special glass that does not grow or shrink with temperature. Only three companies know how to make it well enough.

Technical: Zerodur Li-Al-silicate glass-ceramic, CTE class 0 +/-0.007e-6/K, ~190 blanks/yr sustained; ULE (Corning 7972) titania-silica; Clearceram-Z for TMT. Constraint is ceramisation furnaces and homogeneity metrology, not raw material tonnage.

Frozen prediction: By 2028-12-31 no ELT-class (>=1.4 m, CTE class 0) glass-ceramic blank from a vendor other than SCHOTT, Corning or Ohara is accepted by ESO, TMT or GMT (P=0.9).

HELD modeled (conservative entry; the material ships commercially for decades, but the chokepoint claim rests on USGS material-flow tables with the US production figure withheld). Three ore-to-metal countries (US/Materion, Kazakhstan/Ulba, China); US import sources 2021-24 Kazakhstan 31%, Latvia 25%, Japan 19% (USGS MCS 2025). JWST's 18 O-30 beryllium segments are the application proof.

★ For the kids: Beryllium is a super-light, super-stiff metal that keeps its shape when freezing cold, so the Webb telescope's mirrors are made of it. Only three countries can make it from rock.

Technical: Be: 1.85 g/cm3, E ~300 GPa, CTE near zero below ~100 K; Spor Mountain bertrandite -> Be(OH)2 -> Elmore OH; Ulba sold ~63% via Ulba-China (USGS MYB 2022); world mine output ~325-360 t Be; CBD toxicity slows new entrants.

Frozen prediction: Through 2028-12-31 no beryllium ore-to-metal facility outside the US, Kazakhstan and China reports first production (P=0.85).

HELD modeled under the split rule (bulk LiTaO3/LiNbO3 wafers commercial via Sumitomo Metal Mining and Shin-Etsu for SAW/BAW filters; thin-film LT photonics open - Wang et al., Nature 629, 784, 2024; Nature Photonics 2025). Tantalum's coltan (DRC/Rwanda) exposure is a second-order chokepoint not yet carried.

★ For the kids: Lithium tantalate is a crystal that can switch light on and off billions of times a second. Phone filters already use it; scientists now cut it into thin films for light-based chips.

Technical: LiTaO3 r33 ~30 pm/V, lower birefringence and higher damage threshold than LiNbO3; ion-slice to 600 nm films, 5.6 dB/m loss, 40 GHz modulator, 1.96 V.cm. TFLN modulators commercial; TFLT not.

Frozen prediction: By 2028-12-31 no commercial optical transceiver ships with a thin-film lithium TANTALATE modulator while TFLN modulators remain in shipping products (P=0.7).

HELD open. ~10x lower mechanical loss than amorphous SiO2/Ta2O5 (Cole et al., arXiv 2301.02687; Amato et al., Adv. Photonics Res. 2025), 2.5e-17 Si-cavity stability (arXiv 2509.13503) - but coating diameter capped ~20 cm by GaAs wafer size, plus photo-birefringent (arXiv 2602.04724) and electro-optic (PRD 107, 022003) noise; single-group and preprint-stage.

★ For the kids: Telescopes that listen for ripples in space need mirrors that barely jiggle from heat. Perfect crystal-layer mirrors jiggle ten times less, but nobody can grow them big enough yet.

Technical: MBE GaAs/Al0.92Ga0.08As quarter-wave stacks, epitaxial lift-off, direct bonding to silica/silicon; loss angle ~1e-5 vs ~3e-4 amorphous. Critical-metal link: gallium.

Frozen prediction: By 2028-12-31 no substrate-transferred GaAs/AlGaAs crystalline coating of >=30 cm clear aperture is peer-reviewed or installed on any GW-detector test mass (P=0.75).